All papers
2,340 papers from the Hugging Face Daily Papers feed, each with a podcast summary and its transcript.
Wednesday, November 11, 2026 · 39
- Few-Step Streaming Video Generation via Condition Shifting Tsinghua University, Shengshu Technology, Westlake University · 2611.14033 · ▲ 78
- Tree-Ising: Repairing Language Models with Bit-Level Ising Edits QDeep, ITMO University · 2611.14210 · ▲ 78
- World In-Context Learning for Agentic Robot Manipulation Huazhong University of Science and Technology, ArcheOne AI · 2611.12885 · ▲ 66
- Don't Just Trust Your Memory: Evidence-Linked Memory Reduces Memory-Induced Degradation in LLM Agents KAIST, DeepAuto AI · 2611.14320 · ▲ 56
- dNaViT: Discrete Codes are General-Purpose Visual Representations Meituan's LongCat Team · 2611.12426 · ▲ 33
- MMDM: Measuring Massive Multitask Decision Making Chinese University of Hong Kong, Hong Kong University of Science and Technology, Lightspeed · 2611.14309 · ▲ 26
- DexForesight: Grounding Dexterous Policies with Implicit Spatial Foresight Sharpa Robotics, Shanghai Jiao Tong University, Chinese University of Hong Kong · 2611.11166 · ▲ 24
- Embodied First Exam Northwestern University, Carnegie Mellon University, University of California, Berkeley · 2611.12177 · ▲ 24
- One Table, Many Queries: Benchmarking Generalist Tabular Models Across Generation, Imputation, and Prediction StableAI, Tsinghua University, Beijing University of Posts and Telecommunications · 2611.13001 · ▲ 23
- FINCOSMOS: An Open-World Evaluation Paradigm for Agents in Finance Fin AI, Stevens Institute of Technology, Independent researchers · 2611.07464 · ▲ 22
- MACaQ: Fast, Model Agnostic Calibrated Quantization for Low-Precision Weights and Activations Huawei · 2611.10649 · ▲ 22
- WaveGen: Truly End-to-End Waveform Generation via Internal Spectral Trajectory Learning Ajou University, KT · 2611.12527 · ▲ 15
- DepthGraft: Structural Regularization through Hierarchical Cross-Layer KV Reconstruction Shanghai Jiao Tong University, Shanghai Innovation Institute, Xiaohongshu · 2611.03142 · ▲ 14
- ProTEUS: A mutation-resolved world model for closed-loop protein evolution Foundry Bioscience, Massachusetts Institute of Technology, Stanford University School of Medicine · 2611.13348 · ▲ 13
- From Tool State to Scientific Action: Auditing Decision Provenance in Biomedical Research Agents Intosyn AI Labs, Columbia University, Microsoft · 2611.01527 · ▲ 10
- The Surprising Effectiveness of Deleting Weights in LLM Post-Training New York University, University of Calgary · 2611.10866 · ▲ 10
- PhyRubric: Rubric-Grounded Diagnostic Evaluation of Physical Commonsense in Text-to-Video Generation University of Illinois Chicago · 2611.13576 · ▲ 10
- Test-Time Scaling for Diffusion Language Models with Frequency-Aware Remasking University of California, Riverside, Indiana University · 2611.10486 · ▲ 8
- PhysOPSD: Code as Privilege for Physically Plausible Video Generation Hong Kong University of Science and Technology, Alibaba · 2611.12411 · ▲ 8
- Micius-Flash: Lightweight Agentic Intelligence for Engineering arXiv · 2611.10015 · ▲ 7
- Is Your Robot's Memory Just a Clock? Temporal Shortcuts in Memory-Augmented VLA Policies Nanyang Technological University · 2611.14488 · ▲ 7
- Are Causal Neuron Circuits Unique and Transferable? Iletisim Yazilim · 2611.13885 · ▲ 6
- Market-Conditioned JEPA: Learning from Unpredictable Market Shocks University of Hong Kong, Grace Investment Machine · 2611.13572 · ▲ 5
- LEASH: Stay Safe in Agentic Robotics Xi'an Jiaotong-Liverpool University, Duke Kunshan University, University of Liverpool · 2611.14115 · ▲ 5
- Blind Variation, Selective Retention: Improving Creative Idea Generation in AI Scientists Stanford University · 2611.12184 · ▲ 4
- Shortcuts and Boundaries: Verifiable RL Misalignment and Contextual Safety Learning arXiv · 2611.12865 · ▲ 4
- EarthRecord: From Pixels to Land Change with Multimodal Agents over Four Decades of Satellite Imagery AIEyes Technologies, Hong Kong University of Science and Technology, Guangzhou, Guangdong University of Technology · 2611.12934 · ▲ 4
- ExRL: Sample-Efficient Online Adaptation of Long-Chunk Policies via Execution Length Control University of Electronic Science and Technology of China, Tongji University, Hong Kong Polytechnic University · 2611.13482 · ▲ 4
- Are Dense Rewards Worth Their Tokens in GRPO? Case Western Reserve University · 2611.13560 · ▲ 4
- DyCraft: An Agentic System for Long-Horizon 4D Scene Generation Stanford University, University of Cambridge · 2610.40754 · ▲ 3
- Learning to Act with World Models by Valuing What Comes Next National University of Singapore, University of California, Los Angeles, Fudan University · 2611.12200 · ▲ 3
- Predictive Surprise as Self-Grounding Concept Bottleneck for Interpretable Time Series Nanyang Technological University, A-STAR, Khalifa University · 2611.12955 · ▲ 3
- EVAThink: Evidence Sufficiency-Guided Training for Adaptive Thinking in Video Understanding East China Normal University, Zhongguancun Academy, Huawei · 2611.13907 · ▲ 3
- TorchCurves: Differentiable Parametric Curves arXiv · 2611.14516 · ▲ 3
- Lipschitz-Controlled Flow Matching for Robust Image Restoration Paderborn University, TU Dortmund University · 2611.13995 · ▲ 2
- Sable: A Small-Field Sumcheck Argument for Bit-Exact Inference of Large Language Models MoonMath AI · 2611.14083 · ▲ 2
- Visual Similarity in Adult Word Extension and Artificial Word Learning Neurocreaciones · 2611.14116 · ▲ 2
- LO-World: Unified BEV Latent Modeling for Joint 4D Occupancy and LiDAR Forecasting Xi'an Jiaotong University · 2611.11912 · ▲ 1
- Prompt2Echo (P2Echo): Text-Prompted Multi-View Echocardiography Segmentation with One Model University of British Columbia, EPFL, University of Regensburg · 2611.13425 · ▲ 1
Tuesday, November 10, 2026 · 39
- Learning Heterogeneous Tactile Representation for General Dexterous Manipulation arXiv · 2611.12806 · ▲ 99
- WorldFoundry: Towards Unified Infrastructure for World Intelligence OpenEnvision · 2611.10771 · ▲ 57
- Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery Shanghai AI Laboratory · 2611.12762 · ▲ 54
- Reframe: Observation-Aligned Multi-Teacher On-Policy Distillation for Agentic Visual Reasoning Nanjing University, Institute of Automation, Chinese Academy of Sciences · 2611.12886 · ▲ 39
- Flide: Accelerating Decision Inference with Exact State Forking National University of Singapore, University of Oxford · 2610.42148 · ▲ 35
- Infrathon: Benchmarking Language Agents for Real-World, Long-Horizon AI Infra Optimization University of Washington, Allen Institute for AI, University of Pennsylvania · 2611.11995 · ▲ 35
- InfraRSI: Harnessing Recursive Self-Improvement for Training Infrastructure Peking University · 2611.11558 · ▲ 33
- ReproBridge: Building Reproducibility into ML Research Artifacts University of Waterloo, Google · 2611.08452 · ▲ 31
- SWAM: Scaling World Action Models with Cross-Modal Alignment Shanghai Jiao Tong University, Shanghai Innovation Institute, Noematrix · 2611.12870 · ▲ 22
- Longtail-SWE: High-Quality SWE Tasks for Underrepresented Languages Nebius · 2611.10304 · ▲ 21
- How Many Exposures Does a World Model Need? Learning Unseen Game Mechanics at Test Time Third Intelligence, University of Tokyo · 2611.12324 · ▲ 15
- SIRIS: Can LLMs Tell When Simulation Errors Matter in AI-Driven Systems Research KAIST, Microsoft · 2611.10329 · ▲ 12
- How Much Training-Inference Mismatch Can On-Policy Distillation Tolerate? Aalto University, Nanyang Technological University, Independent researchers · 2611.11450 · ▲ 12
- MotionGraphicsBench: Evaluating Motion Graphics Generation with Programmatic Verification Stanford University, Figma · 2611.10592 · ▲ 11
- FewStepRL: Fine-tuning Few-step Diffusion Models with Reinforcement Learning Tencent, Westlake University, University of Chinese Academy of Sciences · 2611.08240 · ▲ 9
- AutoVend: Benchmarking LLM Agents on Long-Horizon Autonomous Vending Operations StepFun, Zhejiang University, Xi'an Jiaotong-Liverpool University · 2610.36980 · ▲ 8
- Auriga Technical Report: A Task-Agnostic Agent Harness for Zero-Shot Embodied Manipulation StellarEdge AI · 2611.01112 · ▲ 8
- TRIO: Task Refinement with Individual Optimization for Retrieval-Augmented Generation University of Technology Sydney, University of Wisconsin-Madison, Peking University · 2611.02810 · ▲ 8
- e-means: outperforming k-means (in one-tenth the epochs) Meta, École Nationale des Ponts et Chaussées · 2611.09501 · ▲ 8
- All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States University of Southern California · 2611.10123 · ▲ 6
- Training Vision Transformers to Focus: Emergence of Spatial Attention Head Specialization Kioxia · 2611.11890 · ▲ 5
- FlyHash for LLM Agent Memory: An Empirical Validation and a New Application Independent researchers · 2611.05791 · ▲ 4
- SkinMind: Equipping Agentic RAG with Human-Inspired Graph Probing for Reliable Dermatology AI Ant Group, University of Hong Kong, Monash University · 2611.08935 · ▲ 4
- FraudBench: A Legal Evaluation of AI Deception Harvard Law School, University of Oxford, Institute for Law and AI · 2611.12167 · ▲ 4
- LCM-Promo: An LLM-Based Uplift Model for Promotion Allocation Prosus, iFood · 2610.37570 · ▲ 3
- onóna: Twin Embeddings with Tied Unembeddings Independent researchers, Michigan State University, Together AI · 2611.10556 · ▲ 3
- JoVox: Joint Hierarchical Flow Matching for Lightweight Voxel Generation King Abdullah University of Science and Technology · 2611.10690 · ▲ 3
- Modality-Aware Expert Pruning for MoE-Based Multimodal Large Language Models Nota · 2611.05927 · ▲ 2
- PhysioWave-T: Token-Efficient Wavelet Foundation Models for Biosignals Nanyang Technological University, University of Modena and Reggio Emilia, ETH Zurich · 2611.08969 · ▲ 2
- InstantHMR : Real-Time Human Mesh Recovery on Your Phone NaturalPad, LAAS, CNRS · 2611.09195 · ▲ 2
- RidgeFill: Training-Free Late-Layer Cache Fill for Faster Prefill Independent researchers · 2611.09709 · ▲ 2
- Stochastic Deep Supervision for Gallery-Consistent Early-Exit Face Recognition Technical University of Darmstadt, Fraunhofer Institute for Computer Graphics Research · 2611.09987 · ▲ 2
- Spatial Harness: Grounding Spatial Reasoning in Executable World States with Physics Engines Hong Kong Baptist University, ByteDance · 2611.11344 · ▲ 2
- Creative Priming: Decoupling Process from Product to Escape the Creative Paradox Lila Sciences · 2611.11740 · ▲ 2
- EmoHarness: Boosting Multimodal Emotion Understanding with Meta-Harness Evolving University of Illinois Urbana-Champaign, University of California, Davis, Stanford University · 2611.11994 · ▲ 2
- Knowledge Graph-based Retrieval-Augmented Generation for Schema Matching arXiv · 2501.08686 · ▲ 1
- Energy-Tweedie: Score meets Score, Energy meets Energy University of Michigan · 2512.23818 · ▲ 1
- Horizon-Scalable Q-Learning University of Texas at Austin · 2611.10987 · ▲ 1
- Click3R: Interactive Stereo 3D Reconstruction with Sparse Correspondence Clicks University of Virginia · 2611.12875 · ▲ 1
Monday, November 9, 2026 · 33
- Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription NetEase Youdao, University of California San Diego · 2611.06043 · ▲ 91
- SplatAnything: Gaussian Spacing Normalization for Feed-forward 3DGS across Arbitrary Scenes Zhejiang University, Tencent, Causa Robotics · 2611.08264 · ▲ 70
- P^2-MOPD: Accelerating Multi-Teacher On-Policy Distillation via Parameter Priors Adelaide University, StepFun, Shanghai Jiao Tong University · 2611.03953 · ▲ 64
- Toward the Era of Agentic Embodied Intelligence Shanghai Jiao Tong University, Anyverse Dynamics · 2611.09124 · ▲ 53
- PCBRouteAgent: How Close Are LLM Agents to Becoming PCB Routing Engineers? University of Texas at Dallas, Texas Instruments, Iowa State University · 2611.08175 · ▲ 42
- OneTrans-GSR: A Generative Slate Ranking Paradigm Unifying Reranking and Mix-Ranking with One Autoregressive Transformer ByteDance · 2611.04720 · ▲ 40
- LLMs Don't Know What They Know: Benchmarking Temporal Self-Knowledge for Financial Backtesting Fin AI, University of Manchester, Mohamed bin Zayed University of Artificial Intelligence · 2611.06808 · ▲ 33
- LongBanana: An Expert-Verified Benchmark for Long-Context Multi-Reference Image Synthesis South China Normal University, University of British Columbia, Vector Institute · 2611.06856 · ▲ 31
- Frontier3DBench: Benchmarking Coding Agents for 3D Creation across Objects, Scenes, and Interactive Open Worlds Navers Lab, Einsia AI, Tsinghua University · 2611.09017 · ▲ 26
- Beneath the Surface: What We Overlook When We Talk About Agent Evolution University of Science and Technology of China, University of California, Berkeley, Simple Agent Lab · 2611.07624 · ▲ 25
- EvoWorld: Towards Out-of-Sight World Evolution in Generative Video Models Hong Kong University of Science and Technology, Tencent, Chinese University of Hong Kong · 2611.06748 · ▲ 24
- Fast and Exact Relative-Position Attention via Structure-Aware Kernels INSAIT, Sofia University · 2611.01182 · ▲ 22
- AgentBaton: Efficient Model Collaboration in Agentic Tasks NVIDIA, Hong Kong University of Science and Technology, University of Illinois Urbana-Champaign · 2610.35959 · ▲ 21
- Video-Coder: Reinforcing Agentic Coding for Video Generation University of Central Florida, University of Würzburg, Microsoft · 2611.09894 · ▲ 21
- Exploring Weight-Space Programming: Capabilities, Generalization, and Limitations Shanghai AI Laboratory, Tongji University, Fudan University · 2611.09382 · ▲ 18
- LowRAM: Low-Rank Keys and Amortised Values for Composable KV Cache Compression at Test-Time Huawei, University College London · 2611.07025 · ▲ 15
- AutoCompress: Automated Hardware-Guided Model Compression NVIDIA · 2611.08290 · ▲ 11
- Steerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank Adapters University of Sheffield · 2611.09634 · ▲ 11
- CPU-ARC: Agentic Reasoning Caching for on-CPU Reasoning Reuse NVIDIA · 2611.09891 · ▲ 10
- Hierarchical Residuals: A Minimalist Decoupling of Residual Width from Compute Width Tsinghua University, ByteDance · 2610.41711 · ▲ 8
- Authorization Drift in LLM Agents: Evaluating Boundary Compliance in Multi-Step Tool Use Hong Kong University of Science and Technology, South China University of Technology, Suzhou University · 2611.09113 · ▲ 7
- "PRISM" Is Not All You Need: Cross-Topic Name Collisions and Scientific Paper Retrieval Carnegie Mellon University · 2611.01362 · ▲ 6
- Human-Centred, Risk-Constrained Evaluation of Financial Document Recognition: Evidence, Critical Errors, and Selective Automation Independent researchers · 2611.02775 · ▲ 6
- The Last Mile to the Real World: Agent-Native Primitives as Reusable Building Blocks for Solving Real-World Tasks University of Illinois Urbana-Champaign, Georgia Institute of Technology, heelo.dev · 2611.07461 · ▲ 6
- PREEN: Exemplar-Based Color Style Transfer with Smartphone Editor Presets Samsung · 2610.36896 · ▲ 5
- Knowing When to Teach: Learning from Teachable Prefixes in Multilingual Reasoning Distillation KAIST, Korea University · 2611.08179 · ▲ 5
- Coding or Cheating? An Investigation of Frontier Coding LLMs Chinese University of Hong Kong · 2611.10404 · ▲ 5
- Sketchpad: certified exact arithmetic with a small vision-language model MIT Media Lab · 2611.04434 · ▲ 4
- IWIBench: Evaluating Gradient-Free Model Driven Intra-weight Injection HCLTech · 2611.08647 · ▲ 4
- OmniLatent: Unified Latent for 3D Part Modeling University of Hong Kong, Tencent, Chinese University of Hong Kong · 2611.09055 · ▲ 4
- Learning to evolve by matching velocities Politecnico di Milano · 2611.00448 · ▲ 2
- Where to Connect? Boosting MLLMs via Dynamic Gated Pathways across ALL ViT and LLM Layers Northwestern Polytechnical University, Hong Kong University of Science and Technology · 2611.06844 · ▲ 2
- ODE Policy: Robot Learning via Action Dynamics, Not Actions Nanyang Technological University · 2611.01729 · ▲ 1
Friday, November 6, 2026 · 32
- From Autonomous Agents to Cognitive Symbionts: Toward AI-Augmented Human Intelligence University of Oxford · 2611.05376 · ▲ 121
- VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding NVIDIA, Johns Hopkins University, Hong Kong Polytechnic University · 2611.08449 · ▲ 111
- SLA: Scaling Spatial-Language-Action Model via Continual Pre-training Hong Kong University of Science and Technology, Guangzhou, OLA Dimensions, Beijing Academy of Artificial Intelligence · 2611.08373 · ▲ 99
- Harness-S: Learning from Sibling Harnesses for Generalizable Agent Training Shanghai Jiao Tong University, Xiaohongshu, Zhongguancun Institute of Artificial Intelligence · 2611.04277 · ▲ 67
- Random Is Almost Enough? Stable and Low-Cost Benchmark Compression via BestRand Bosch Research, Carnegie Mellon University · 2611.00883 · ▲ 53
- WeLM-Image: A Unified Image Generation Model towards Visual Intelligence Tencent · 2611.07392 · ▲ 48
- TailSteer: Trajectory-Tail Steering for Drift-Resistant Multi-turn Image Editing Chinese University of Hong Kong, SenseTime, Shanghai AI Laboratory · 2611.06741 · ▲ 43
- Open-HumanScale: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models Peking University, National University of Singapore, Massachusetts Institute of Technology · 2611.06817 · ▲ 39
- EvoMemState: Evolving Memory State for Personal Agents Xiamen University, Nanyang Technological University, University of British Columbia · 2611.03743 · ▲ 35
- Towards Foundation Decision Models: Drex Nace AI · 2611.07854 · ▲ 24
- VisSearch-Bench: Evaluating Long Horizon Visual Search in Multimodal Agents Tsinghua University, Ant Group, Renmin University of China · 2611.04894 · ▲ 22
- EDGAR: Efficient Diffusion-Driven Gaussian Avatar Refinement via Synthetic Failure KAIST · 2611.07909 · ▲ 18
- On the Opportunities and Challenges of Prefill-to-Autoregressive Cross-Model Transfer King's College London, Tencent, National Physical Laboratory · 2611.06976 · ▲ 15
- Muir: Linear Attention with Sensitivity-Allocated Gating and Matched-Filter Readout University of Texas at Austin, Google · 2611.08376 · ▲ 13
- Terminal-style Mathematical Modeling Bench: The First Agentic and Verifiable Benchmark for Mathematical Modeling IDEA Research, Hong Kong University of Science and Technology, Guangzhou, DataArcTech · 2611.08237 · ▲ 12
- Future Representation Distillation for Imagination-free World Action Models KAIST · 2611.04527 · ▲ 11
- SwarmEvolve: Latent Agent Swarms for Evolutionary Discovery and Optimization Brown University · 2611.08483 · ▲ 11
- HeLM: Hierarchical Explicit Language Memory for Non-Markovian Manipulation Seoul National University · 2611.02660 · ▲ 9
- Decode Before You Hear: Speculative Listening for Chunk-Interleaved Full-Duplex Speech Language Models Chinese University of Hong Kong, Shenzhen, Amphion Technology · 2611.06982 · ▲ 9
- SAE Interprets MoE: Sparse AutoEncoders Reveal That MoE Experts Exhibit Human-Interpretable Fine-Grained Specialization ByteDance, Shanghai University of Finance and Economics · 2611.08089 · ▲ 8
- OpenFrontiers: A Collection of Open Research Problems and a Benchmark for Scientific Rediscovery University of Illinois Urbana-Champaign · 2610.37739 · ▲ 7
- Tools Need Not Speak in Text: Latent Context Compression for Agents Tsinghua University · 2611.06500 · ▲ 7
- Reducing Representational Interference in Retrieval Embeddings with Trainable Prefix Tokens Korea University, Soongsil University · 2611.07298 · ▲ 7
- Decoupled Predictive Decision Making: A Shared Conditional Predictor for Agents Stable AI, Tsinghua University · 2611.02075 · ▲ 5
- StepX-OPD: Rethinking Multi-Teacher On-Policy Distillation in Diffusion Models StepFun, Westlake University · 2611.04386 · ▲ 5
- Signed but Steerable: Measuring and Closing the Text-Vector Integrity Gap in Offline Retrieval-Augmented Generation Securagen AI · 2611.06498 · ▲ 5
- Towards Real-Time Full-Waveform LiDAR Transformers via Intensity-Guided Token Reduction and Physics-Aware Augmentation Keio University · 2611.07613 · ▲ 5
- Identically Trained Language Models Break on Different Jailbreaks Cerulean Labs, National Taiwan University, National Yang Ming Chiao Tung University · 2610.34074 · ▲ 4
- Information Games: Strategic Crowding and Firm Repositioning in Language-Model Space University of Cape Town · 2609.37820 · ▲ 3
- NPCBench: A Clinical Apprenticeship Benchmark for Multimodal Care-Pathway Reasoning in Nasopharyngeal Carcinoma Hong Kong Polytechnic University, Infix AI, Sun Yat-sen University · 2611.08369 · ▲ 3
- Hard Sample-Conditioned Decorrelated Attention for Robust Face Recognition Fraunhofer Institute for Computer Graphics Research, TU Darmstadt · 2610.41789 · ▲ 2
- SynthFaces: Balanced Large Scale Human Dataset FusionBrain Lab, HSE University, Moscow State University · 2610.40119 · ▲ 1
Thursday, November 5, 2026 · 39
- From Pixels to Actions - Towards Native Vision-Language-Action Policies Nanyang Technological University, SenseTime, Ropedia · 2611.06918 · ▲ 62
- FrontierPhysics: Evaluating Agents for End-to-End Physics Research BenchFlow, University of California, Berkeley, California Institute of Technology · 2611.06910 · ▲ 59
- Turns Matter: Advancing Streaming Video RL via Turn-Level Credit Fudan University, Shanghai Innovation Institute, JD.com · 2611.06728 · ▲ 51
- What Does Multi-Turn Reinforcement Learning Really Improve in LLM Agents? Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Tsinghua University · 2611.06483 · ▲ 39
- GameVerse: A Minute-Scale Gameplay Dataset for Long-Horizon Interactive World Modeling Alaya Lab, Shanghai Innovation Institute, Wuhan University · 2610.37257 · ▲ 37
- Halley: A Predictive Control Plane for Optical AI Data Center Networks Prompeteer AI, Indian Institute of Technology Madras · 2611.04408 · ▲ 33 · Deep dive
- VideoMemBench: How Well Do MLLMs Remember in Video Conversations? University of Technology Sydney, Chinese University of Hong Kong, Hong Kong University of Science and Technology · 2610.40131 · ▲ 31
- EmoReward: Self-Evolving Reward Calibration for Emotional Video Generation RL Hong Kong University of Science and Technology, Tencent, Fudan University · 2611.06518 · ▲ 27
- PyRUA-DataFarm: LLM Agent as a Cost-Efficient Teleoperator for Real-Ready Robot Data Peking University, NVIDIA · 2611.05320 · ▲ 25
- On the Generalization Limitation of On-Policy Distillation for Post-RL Consolidation University of Chicago, Duke University, University of Illinois Urbana-Champaign · 2611.05609 · ▲ 24
- StreamDuo: A Live Commentator and a Research Agent for Unseen Games Alaya Lab, University of Tokyo, Georgia Institute of Technology · 2611.01120 · ▲ 22 · Deep dive
- MUON-RT Reusing the Teacher's Trajectory for Momentum-Free Muon for LLM Self-Distillation Indian Institute of Technology Kanpur, Amazon · 2610.41104 · ▲ 22
- D3D: Decoupled Diffusion Drafting for Asynchronous Speculative Decoding Shanghai Innovation Institute, Shanghai Academy of AI for Science, Fudan University · 2611.06425 · ▲ 19
- Solv3R: 3D Looped Transformer as a Multi-view Solver University of British Columbia, Wayve Labs, Simon Fraser University · 2611.02857 · ▲ 18
- Hybrid Computer Use: A Unified Programmatic, Semantic, and Visual Interface for General Agents Microsoft, Fudan University, Tsinghua University · 2610.41432 · ▲ 17 · Deep dive
- TAC-CoT: Tool-Empowered Implicit Chain-of-Thought via Interleaving Explicit Tokens KAIST, Cluvion · 2611.01926 · ▲ 17 · Deep dive
- Human-to-Robot Transfer via Bridging Representation and Action Space in VLAs KAIST, RLWRLD · 2611.04391 · ▲ 16
- SpecOPD: Branch-Parallel On-Policy Distillation for Speculative Decoding Alibaba, University of Science and Technology of China, Zhejiang University · 2611.02846 · ▲ 15
- Why Pay for Long Reasoning When Rewrite Do Trick? Causal Effects of Reasoning History Compaction on Future Generation and Agent Cost Technological School Electronic Systems · 2611.03852 · ▲ 13
- VA-TDM: A Unified View for Improving Few-Step Diffusion Models by Velocity Adversarial Distillation Hong Kong University of Science and Technology, Chinese University of Hong Kong, Shenzhen · 2611.06796 · ▲ 12
- LoopVQ: Sub 1-bit KV Cache Quantization for Looped Transformers Seoul National University · 2611.05304 · ▲ 11 · Deep dive
- AnchorComplete: Recovering Complete Scene Meshes from Partial Point Clouds Technical University of Munich, Munich Center for Machine Learning, Nanyang Technological University · 2611.03228 · ▲ 11
- PyRUA-REAL: Robot Exploration and Adaptive Learning through RSI in the Real World Peking University, EngineAI, National University of Singapore · 2611.06768 · ▲ 11
- The SIMPLE Law for Energy Cost of LLM Inference University of Bologna, CINECA · 2611.03330 · ▲ 9 · Deep dive
- VidDub: Replacement-Aware Distillation for Two-Step Video Dubbing Hong Kong University of Science and Technology, Tencent, Zhejiang University · 2611.03404 · ▲ 9
- How I Aligned Your Model: Scaling Up Multilingual Safety Training with Templates ETH Zurich · 2610.40516 · ▲ 8
- NitroFusion2: Bringing Adversarial Loss to the Text Stream of Multimodal Diffusion Transformers University of Surrey, Turing Elite Research Labs · 2611.05648 · ▲ 8
- LiveFoley: Dynamic Real-Time Video-to-Audio Generation for Immersive Interactive Worlds Monash University, Nanjing University, ShengShu Technology · 2611.05988 · ▲ 8
- SleuthVLM: Do We Really Need Audio-in-Video for Omni Video Understanding? University of Bristol, Ant Group, Tsinghua University · 2611.06083 · ▲ 7
- Multi-layer Indexer: Overlapping Selection and Attention in DeepSeek Sparse Attention Writer · 2611.03768 · ▲ 5 · Deep dive
- VeriSpoof: Beyond Real or Fake with One Audio LLM University of Melbourne · 2611.04634 · ▲ 5
- Monitor Accuracy Isn't the Answer: A Decision-Value Framework for Evaluating AI Agent Restart Monitors Intuit · 2610.36544 · ▲ 4 · Deep dive
- Ask, Don't Equip: Question-Driven Coding Agents for Deep Kernel Optimization Snowflake AI Research, Carnegie Mellon University · 2611.03517 · ▲ 4
- Koopman Generative Operators for Efficient Probabilistic Time-Series Forecasting Ben-Gurion University of the Negev, Amazon · 2611.04760 · ▲ 3
- GaussianRenderFormer: Transformer-based Rendering of 3D Gaussian Splats Hebrew University of Jerusalem · 2611.05593 · ▲ 3
- TaoMate-HD: Progressive-Resolution Autoregressive Audio-Video Generation Alibaba, Nanjing University · 2611.03394 · ▲ 2
- UniRestorer++: Scalable Universal Image Restoration with Multi-Granularity Degradation Modeling arXiv · 2611.06042 · ▲ 2
- HERIT: Enabling Global Access to Hanja-to-English Historical Translation by Mitigating Temporal Bias Kangwon National University · 2611.04949 · ▲ 1
- DRAP: Dynamics-Risk Adaptive Plasticity for Continual Learning in Physical AI arXiv · 2611.05029
Wednesday, November 4, 2026 · 52
- Route Anything: A Study of Model and Policy Routing for General-Purpose AI Systems National University of Singapore, InfRec, Cardinal AI Lab, University of California, Berkeley · 2611.05275 · ▲ 79
- Words Caught in the Heart: Simulating Person-Specific Expression for Self-Opaque Clients in Psychotherapy Wuhan University, Cognitive Language and Information Lab, Huazhong Agricultural University · 2611.00617 · ▲ 77
- MoviEdit: Long-form Multi-Shot Video Editing Made Easy Adobe, KAIST · 2611.02737 · ▲ 74
- Behind Video Generation: Observations and Insights on Multimodal Understanding from Kling Kuaishou · 2611.04916 · ▲ 74
- Think Less, Locally: Local Preference Learning for Reasoning Compression Xi'an Jiaotong University, Tencent · 2611.04814 · ▲ 61
- DreamX: Video Diffusion Model is a Universal Latent Action Learner KAIST, NVIDIA · 2611.05377 · ▲ 57
- RoboColosseum: Evaluating Physical Intelligence at Scale In-The-Wild National University of Singapore, University of Washington, Stanford University · 2611.02497 · ▲ 50
- Axion 1.0: Open Agent Foundation Models, Simplified Duke University, ByteDance, University of Maryland · 2611.03233 · ▲ 48
- Factorized MLLM Embeddings for Composed Image Retrieval KAIST, Google · 2611.05193 · ▲ 48
- ToolSpace: Scaling Tool-Centric Environment Synthesis for Long-Horizon Agents Renmin University of China, Fudan University, Shanghai Jiao Tong University · 2611.05036 · ▲ 41
- YANAMI: A Large-Scale Multimodal Dataset and Benchmark for Multiturn Stateful Interaction Hong Kong University of Science and Technology, Independent researchers, University of Melbourne · 2611.05280 · ▲ 31
- EvoTeam: Multi-Agent Collaboration via Evolving Teams Jilin University, Nanjing University of Science and Technology, Shanghai AI Laboratory · 2610.39813 · ▲ 27 · Deep dive
- WebHarbor: Docking Real Websites for Evolving GUI Agent Environments arXiv · 2611.03634 · ▲ 27
- VidUEU-Agent: A Data-Curation Agent for Multimodal Understanding, Editing, and Unified Tasks Zhejiang University, University of Hong Kong, Institute of Automation, Chinese Academy of Sciences · 2611.04418 · ▲ 27
- Are Frontier VLM Agents Ready to Be Robot Generalists? An Empirical Study with the Embodied Agent Arena Hong Kong University of Science and Technology, Guangzhou, Chinese University of Hong Kong, Knowin AI · 2610.00854 · ▲ 22
- Mage-VL-Cowork: A Full Training Recipe to Learn Hybrid Computer Use From Evolved Harnesses Microsoft · 2610.40747 · ▲ 19 · Deep dive
- ParetoSaddler: Pareto-Guided Optimization of Agent Harnesses for Quality and Cost University College London, POSTECH, Southern University of Science and Technology · 2611.02079 · ▲ 19 · Deep dive
- SpecPilot: Autonomous Algorithm Discovery and Systems Optimization for Speculative Decoding University of California, Berkeley, Princeton University, Stanford University · 2611.02501 · ▲ 19
- MoLRSI: A Recursive Self-Improving AI agent for Molecular Discovery Northeastern University, Broad Institute of MIT and Harvard, Harvard University · 2610.39714 · ▲ 18
- You don't need long context arXiv · 2611.05306 · ▲ 18
- Recursa: Toward Recursively Self-Improving Foundation Model Systems Alibaba · 2611.04035 · ▲ 16
- "Shall We Approve?" Consequence-Aware Approval for Autonomous Coding Agents Singapore Management University · 2610.40753 · ▲ 15
- Every Token Needs Its Own Teacher: Multi-teacher Adaptive On-policy Self-distillation Technical University of Munich, Munich Center for Machine Learning, Konrad Zuse School of Excellence in Reliable AI · 2610.40574 · ▲ 13 · Deep dive
- Dory1.0-4B: A Looped Normalized-GPT arXiv · 2611.03998 · ▲ 12 · Deep dive
- RoboTryOn: Synthesizing Dexterous Robot Demonstrations from Human Videos via Virtual Try-On KAIST, RLWRLD · 2611.05163 · ▲ 12
- Can AI Reviewers See Your Efforts to Improve Your Paper? A Large-Scale Study of Paper Revision Trajectories Mohamed bin Zayed University of Artificial Intelligence, Virginia Tech, University of Maryland · 2611.02244 · ▲ 11
- FlowTranslator: Training-Free In-Image Machine Translation with Flow-Based Editing Fudan University, Shanghai Key Laboratory of Multimodal Embodied AI · 2611.04303 · ▲ 11
- Switch-SWE: Do Coding Agents Hold Focus When Switching Between Unfinished Tasks? Shanghai Jiao Tong University, Tongji University · 2611.02017 · ▲ 10
- PrimeSim: A Generalist World Model Simulator from Large-Scale Cross-Embodiment Demonstrations PrimeBot Research, Southeast University, Shanghai Jiao Tong University · 2611.04870 · ▲ 10
- Adversarial Speculative Reasoning Improves Long Horizon Agents Subconscious Systems Technologies, Massachusetts Institute of Technology, Princeton University · 2611.05345 · ▲ 9
- LOLPlay: Benchmarking Real-Time Visual Agents in League of Legends arXiv · 2611.03685 · ▲ 8
- Near-Zero Token Embeddings Silently Stall Training in Gated Sequence Models arXiv · 2611.04449 · ▲ 7
- Do We Need an LLM to Know Why Sources Conflict? Decision Models for Knowledge-Conflict Typing in RAG Independent researchers · 2610.36915 · ▲ 5
- SeMo3D: Unifying 3D Generation, Editing, and Completion via Selective Modeling Nanjing University · 2611.02568 · ▲ 5
- Muon Makes the First Mixture-of-Experts Layer Harder to Balance Autonomous University of Madrid · 2610.37914 · ▲ 4
- Is Your Long-Horizon Benchmark Really Deep? Measuring and Synthesizing Depth in Terminal Tasks University of Notre Dame, IBM, Carnegie Mellon University · 2611.01927 · ▲ 4
- Is Forgetting Unified? Exploring Concept Unlearning in Unified Multimodal Models Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences, ByteDance · 2611.04467 · ▲ 4
- Going Deeper With Memories University of Luxembourg · 2611.05255 · ▲ 4
- Where to Learn and How Far to Roll Out: Efficient Adaptive Training for GUI Agents Alibaba, Chinese University of Hong Kong · 2610.39251 · ▲ 3
- What We Keep of What We Forget: Replacement Operators and Routing Semantics in Training-Free MoE Expert Pruning National University of Sciences and Technology, National Center of Artificial Intelligence · 2610.41366 · ▲ 3
- BEND 2.0: Benchmarking the Capabilities Needed to Design DNA University of Copenhagen, Technical University of Denmark, Novo Nordisk · 2611.03806 · ▲ 3
- Cross-Prompt Reuse of Entity-Replacement Responses Dalhousie University, Meta · 2611.00352 · ▲ 2
- FinForesight: A Diagnostic Backtesting Benchmark for Deep Research Systems in Financial Forecasting University of Hong Kong, Hong Kong Polytechnic University, University of Cambridge · 2611.01621 · ▲ 2
- GigaSpeech 3: Scaling Multi-Dialect Arabic Speech Recognition with Weakly Labeled Data Shanghai Jiao Tong University, Shanghai Innovation Institute, University of Chinese Academy of Sciences · 2611.02476 · ▲ 2
- Auto-Research Agent on MSA Organization to Improve Protein-Protein Complex Structure Prediction Beijing University of Posts and Telecommunications, Shanghai Jiao Tong University, Fudan University · 2611.03183 · ▲ 2
- Every Pixel Is an Option: A System One Model for Medical Ultrasound Biometry University of British Columbia, Rutgers Cancer Institute · 2611.03360 · ▲ 2
- Enhancing Stock Market Information Retrieval with Retrieval-Augmented Generation FPT University · 2611.03747 · ▲ 2
- Kierston & KierstonBench: Simulation of Allentown, PA and Evaluation of Language Models in Crime Investigation and Surveillance Tasks arXiv · 2611.04785 · ▲ 1 · Deep dive
- QFabric: Guarded Cross-Domain Task Recovery and Its Limits on a Linux-RTOS Board Pollinations AI, JIS University · 2611.01598 · ▲ 1
- SphereFlow: Missing Modality Imputation via Geometric Transport on Hypersphere Hong Kong Polytechnic University, Zhejiang University, IROOTECH Technology · 2611.04414 · ▲ 1
- Say the Same, Act Differently: Diagnosing and Steering Text-Action Divergence in Reasoning VLAs University of Southern California, University of Arizona, General Motors · 2611.04166
- Length Is Cheap, Competition Is Not: Attention Predicts Context Rot in Retrieval University of California San Diego · 2611.04169
Tuesday, November 3, 2026 · 45
- Advances in Agentic Computer Graphics University of Texas at Austin, Stanford University, Simon Fraser University · 2611.01860 · ▲ 143
- Wave Forcing: Towards Speed-of-Light Streaming Video Generation Shanghai Jiao Tong University, Xi'an Jiaotong University, Peking University · 2611.00859 · ▲ 126 · Deep dive
- CoworkOS:Scaling Agent Coworkers for Real-world Automation arXiv · 2611.03178 · ▲ 101
- Video Generation Meets Agent Harness: Agentic Video Generation, Agentic World Modeling and Recursive Self-Improvement Video Rebirth, National University of Singapore, Nanyang Technological University · 2610.40191 · ▲ 78
- Democratizing Robot-Use Agents: A Unified Harness with Open Models Tsinghua University, University of California, Berkeley · 2611.01560 · ▲ 48 · Deep dive
- CUA-Master: Scaling Computer-use Agents through Open-World Recursive Improvements Chinese University of Hong Kong, Tsinghua University, Hong Kong University of Science and Technology, Guangzhou · 2611.03126 · ▲ 47
- IdeaTaste: Do LLMs Share the Research Taste of Human Researchers? Carnegie Mellon University, Mohamed bin Zayed University of Artificial Intelligence · 2611.03424 · ▲ 47
- SWE-Trail: History-Guided Trajectory Collection for Coding Agent Training University of Connecticut, Ant Group · 2611.02074 · ▲ 40
- PhotoGuider: Composition-Aware Cropping and 360^{circ} View Framing with Explainable Photographic Assessments MediaTek, National Taiwan University, National Yang Ming Chiao Tung University · 2611.02897 · ▲ 38
- Lost in Recovery: Demystifying Recovery Behavior of Computer-Use Agents via Goal-Oriented Action Replay KAIST, ETRI, DeepAuto.ai · 2610.27954 · ▲ 35
- Order-Only On-Policy Distillation University of Memphis, University of Missouri · 2611.02112 · ▲ 34
- NorthCUA Technical Report: Turbocharging Computer-Use Agents through Experience at Scale Shanghai Qiji Zhifeng · 2611.02102 · ▲ 32 · Deep dive
- CrossSWE: Benchmarking Cross-Repository Joint Development for Coding Agents Shanghai Jiao Tong University, Xidian University, Singapore Management University · 2611.01484 · ▲ 25
- Reviso: Not All Visual Evidence Deserves Equal Attention in Video Captioning Peking University, KlingAI Research, Hong Kong University of Science and Technology, Guangzhou · 2611.03602 · ▲ 25
- When Does Latent Prediction Error Support Physical Surprise Claims in Video World Models? Technology School Electronic Systems, Baba Tonka High School of Mathematics · 2611.01815 · ▲ 24
- From Fragmentation to Fusion: A Survey and Roadmap for Omnimodal Embeddings Ant Group, Shanghai Jiao Tong University, Australian National University · 2611.00172 · ▲ 23
- History Forcing: Distilling History-Dependent Continuations for Long-Horizon Video World Models Zhejiang University, Microsoft, Monash University · 2611.02782 · ▲ 19
- Diagnosing the Reference Gap: What Visual Scaffolds Can and Cannot Fix in Multimodal Reasoning Nanyang Technological University, Alibaba, University of Macau · 2610.39419 · ▲ 18
- FLAMIA: FLAsh MIcroscaling Attention for Fast and Causally Safe Training on Blackwell Yandex Research · 2611.03344 · ▲ 18
- Avatar Intelligence: Amplifying Human Agency Through Persistent Personal Agents Shanghai Jiao Tong University, National University of Singapore, Shanghai University · 2611.03092 · ▲ 16 · Deep dive
- Measuring AI Agents as Teammates with Collaborative Intelligence Stanford University, Massachusetts Institute of Technology, University of Oxford · 2611.02633 · ▲ 15 · Deep dive
- DecisionRank: On-Policy Distillation for Decision-Model Rerankers Meta, Purdue University · 2610.42128 · ▲ 15
- DeFT-RAG: Decomposing Multi-Hop Questions into Self-Contained Sub-Questions with Reusable Answers University of Texas at Dallas · 2610.37826 · ▲ 13
- LEANGUARD: Lean-Verified Guardrails for Tool-Using Agents Case Western Reserve University, University of Texas at Austin · 2611.02332 · ▲ 12
- Library of Layers (LoL): Adaptively Ordering Transformer Layers at Runtime EPFL, University of Maryland, New York University · 2611.03526 · ▲ 11 · Deep dive
- Cross-Primitive Consistency in Jev Zhejiang University, vivo · 2610.38623 · ▲ 10 · Deep dive
- BYOD: Build Your Own DLM - Efficient Autoregressive-to-Diffusion Language Model Conversion University Medical Center Utrecht, Utrecht University · 2610.41574 · ▲ 10 · Deep dive
- MAServeBench: Benchmarking Multi-Agent LLM Serving Systems with Hierarchical Collaboration Workloads Peking University, Georgia Institute of Technology · 2610.40293 · ▲ 10
- Multi-Agent Relay Thinking Zhejiang University, Nanyang Technological University, University of Hong Kong · 2610.35219 · ▲ 9
- Self-Improvement with Auxiliary Rewards: Learning from Verified Traces and Evaluating Transfer Georgia Institute of Technology · 2610.41296 · ▲ 8
- On-Policy Self-Distillation with Teacher Evolution Institute of Computing Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Meituan · 2610.38637 · ▲ 7 · Deep dive
- ProVe: Proposer-Verifier Decoding Nanyang Technological University, ByteDance · 2610.27815 · ▲ 7
- Can Coding Agents Sustain Continual Evolution without Forgetting? Tsinghua University · 2611.03219 · ▲ 7
- DR Pixel: Page-Native Deep Research with Vision-Language Models NYU Shanghai, University of California, Berkeley, Rational Dynamics · 2611.03097 · ▲ 6
- Advancing SVD-based LLM Compression via Layer-Wise Error Model Search Technical University of Munich, BMW Group, University of Glasgow · 2610.32878 · ▲ 5 · Deep dive
- DAMI: Dual-Axis Memory Initialization for Enhanced Long-Context Retrieval in Gated Linear Attention SpeedAI, Beihang University · 2610.41994 · ▲ 5 · Deep dive
- WorldFlow: Joint Video and 3D Motion Prediction via Large-Scale Pretraining RoboScience, National University of Singapore · 2610.39278 · ▲ 5
- SIMIT: Self-Improving Vision-Language Models via Imagination at Test-Time Technical University of Denmark, Pioneer Center for AI, University of Copenhagen · 2611.00053 · ▲ 5
- Can Large Language Model Agents Learn by Watching Tutorial Videos? University College London · 2611.01460 · ▲ 5
- A Tutorial on Generative Models for Music Audio Warsaw University of Technology, ZAiKS Lab, Singapore University of Technology and Design · 2610.38252 · ▲ 4 · Deep dive
- TinyAstral: Transferring Self-Evolving Harnesses to Tiny Generalist Manipulation Agents Huawei, Independent researchers · 2610.41216 · ▲ 4 · Deep dive
- Where Do Vision Language Models Fail at 3D Reasoning? Novel, University of Toronto, Vector Institute · 2610.39639 · ▲ 4
- Lab2Sim: Turning Real-World Experiment Videos into Interactive Virtual Experiments South China University of Technology, Hong Kong Polytechnic University · 2611.02928 · ▲ 4
- TaoMate-FVSR:Audio-Guided Facial Video Restoration Alibaba · 2611.03393 · ▲ 2
- When Morphs Look Better: Face Image Quality as a Training-Free Signal for Morphing Attack Detection Technical University of Darmstadt, Fraunhofer Institute for Computer Graphics Research · 2611.00074
Monday, November 2, 2026 · 38
- Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute? Tsinghua University · 2610.39497 · ▲ 64
- R^2Orch: Recursive Self-Improvement for Agent Orchestration via Self-Play within a Skill-Internalization Loop Chinese University of Hong Kong, Shenzhen, Fudan University, University of Illinois Urbana-Champaign · 2610.22740 · ▲ 46
- Latent Reasoning in Continuous Space for Unified Multimodal Models KAIST, POSTECH, RLWRLD · 2610.42149 · ▲ 44 · Deep dive
- Apriel Voice 27B Preview: An Open Model for Speech, Reasoning, and Tool Use ServiceNow · 2610.40738 · ▲ 37 · Deep dive
- FrameForge: Harnessing Priors of Video Foundation Models for Unified Image Generation University of Science and Technology of China, FrameX AI · 2610.36545 · ▲ 34
- Follow the Gradient: Tracing Safety Degradation Back to Training Data Datology AI, University of Illinois Urbana-Champaign · 2610.37481 · ▲ 32 · Deep dive
- CEPO: A General Framework for Improving Prompt Optimization in Multi-Agent Systems Korea University, KAIST, Google · 2610.34722 · ▲ 27
- Wizard: Manipulate Any World in Real Time via Agentic Control Tencent, National University of Singapore, Nanyang Technological University · 2610.40339 · ▲ 27
- ReLift: Dense 3D Tracking as Lifted Trajectory Repair Technion, Israel Institute of Technology, NVIDIA · 2610.42136 · ▲ 27
- HiFOR: Hierarchical Foresight with Action-Guided World Grounding for Vision-Language-Action Model Nanjing University of Science and Technology, National University of Singapore, Shenyang Institute of Automation of the Chinese Academy of Sciences · 2610.40373 · ▲ 24
- Venus-DefakerOne2: A Vision-Forensics Foundation Model with Broader Coverage Ant Group, China Society of Image and Graphics · 2610.37918 · ▲ 21
- CaPtain: A Pipeline for Post-Training Code-as-Policy Robot Agents Chinese University of Hong Kong, NovaXBot · 2610.40799 · ▲ 19 · Deep dive
- Probing Spatial State Representations in Visual Foundation Models New York University, Nanyang Technological University, Hong Kong University of Science and Technology · 2610.40636 · ▲ 19
- REA: A Routed Embodied Agent Runtime for Asynchronous Brain-Cerebellum Control National University of Singapore, InfRec's Cardinal AI Lab, University of California, Berkeley · 2610.42053 · ▲ 16 · Deep dive
- GLOW-FM: Efficient Generative Low-Light Image Enhancement with One-Step Latent Flow Matching Arm AI Research · 2610.31222 · ▲ 13
- WorldBridge: Bridging Video and Action for Task Generalization in World Action Models Joy Future Academy · 2610.40748 · ▲ 13
- When Requirements Interact: Scaling Long-Horizon Coding via Interdependent Specifications Xiaomi, Peking University, University of Hong Kong · 2610.40256 · ▲ 11
- On the Origin and Evolution of Training-Inference Mismatch in LLM Reinforcement Learning George Washington University, Duke University · 2610.40945 · ▲ 10
- SOMA: A Benchmark for Fine-Grained Egocentric Behavior and Body-Motion Understanding EPFL · 2610.36384 · ▲ 9
- HALO: Human-Grounded Granular Evaluation of Proactive Agents University of California, Davis, twenty seventy-seven AI, University of Southern California · 2610.39086 · ▲ 9
- RoboArchitect: Automating the Design of Robotic Agents with Persistent Memory Stanford University · 2610.42141 · ▲ 8 · Deep dive
- OSWAM: An Effective World Action Model with One-Step Action Generation Hong Kong Polytechnic University · 2610.35850 · ▲ 8
- GEPA-Curator: Curated Learning for Self Improving Agents University of California, Berkeley, International Computer Science Institute, Lawrence Berkeley National Laboratory · 2610.37651 · ▲ 8
- MemSteer: Turning User Histories into Supervision for Agent Memory Optimization Peking University, Microsoft · 2610.39276 · ▲ 8
- No Channels Left Behind: Understanding Diffusion Acceleration through Residual Gates Yonsei University, NAVER AI Lab · 2610.40914 · ▲ 7
- Supervising the Visible State: Separating Representation from View Selection in Streaming Video Westlake University, Zhejiang University, University of Science and Technology of China · 2610.34926 · ▲ 6
- WeFair: Fidelity-Aware Alignment from Human Feedback for Personalized Query Rewriting in Short-Video Search Tencent, East China Normal University · 2610.16927 · ▲ 5
- Think Twice: The Metis-1.6 Technical Report Lernex Research · 2610.37903 · ▲ 5
- VoxLock: Target-Speaker Extraction for Voice Agents Without an Enrollment Step Independent researchers · 2610.40727 · ▲ 4 · Deep dive
- CausalWAM: Learning the Causes of Actions in World Action Models University of Illinois Urbana-Champaign, University of California San Diego · 2610.42096 · ▲ 4
- TernScope: Model-Hardware Co-Design for Efficient Ternary Endoscopic Segmentation Monash University Malaysia, Hong Kong University of Science and Technology, Shenzhen Institutes of Advanced Technology of the Chinese Academy of Sciences · 2610.29617 · ▲ 3
- SciGame: Let Agents Play the Scientific Discovery Games in the Real and Counterfactual World Mohamed bin Zayed University of Artificial Intelligence, McGill University, Shanghai AI Laboratory · 2610.37919 · ▲ 3
- rTaste: Do Research Agents Have Taste? Evaluating Early Stage Research Judgment University of Minnesota · 2610.40488 · ▲ 3
- Learning Quality-Controllable Robot Policies with Iterative Hindsight Relabeling KAIST, LG Electronics, Microsoft · 2610.37393 · ▲ 2
- TSFLab: A Self-Improving Agent-Agnostic Workbench for Time Series Forecasting Research University of Houston, University of Illinois Urbana-Champaign, Georgia Institute of Technology · 2610.40901 · ▲ 2
- SketchDirector: Complex Scene Sketching via Agentic Layout Reasoning and Detail Refinement Great Bay University · 2610.41260 · ▲ 2
- TurnTurk-GPT: Text-Based Turn-Taking Prediction for Turkish Ata Technology Platforms · 2610.42010 · ▲ 2
- AuditFL: A Contract-Backed Clearing Layer for Open Federated Learning Hong Kong University of Science and Technology, University of Thessaly · 2610.41278 · ▲ 1 · Deep dive
Friday, October 30, 2026 · 39
- Look Only When Needed: Certificate-Guided Visual Routing for Efficient Diffusion VLMs KAIST, AITRICS · 2610.36671 · ▲ 125
- Your Benchmark Is an Empirical Measure Over Difficulty University of Illinois Urbana-Champaign, University of California San Diego, New York University · 2610.29927 · ▲ 112
- A-RSI: Adaptive Task Refresh for Zero-Data Multimodal Self-Improvement Institute of Information Engineering of the Chinese Academy of Sciences, Chinese University of Hong Kong, University of Würzburg · 2610.38568 · ▲ 85
- Proving and Improving the Platonic Representation Hypothesis with JEPAs Brown University, New York University, AMI Labs · 2610.39302 · ▲ 72
- WorkFactory: Turning Professional Intents into Executable Work for Scalable Agent Training IQuest Research, Zhejiang University, Renmin University of China · 2610.37161 · ▲ 64
- L3G-World: Real-Time Infinite World Generation with Latent 3D Gaussian Memory KAIST, Ewha Womans University, KAIST Urban AI Institute · 2610.40614 · ▲ 52
- A Broader Look into MLLM's Retrieval Heads Institute of Information Engineering of the Chinese Academy of Sciences, University of Chinese Academy of Sciences, Tencent · 2610.32411 · ▲ 43 · Deep dive
- HairDeck: Animation-Ready Stylized Hair Cards from a Single Image Tsinghua University, Carnegie Mellon University, Chinese University of Hong Kong · 2610.40599 · ▲ 41
- VibeGame: Prompt-to-Game Development with AI-Native Engine and Self-Evolving Adversarial Agent Team Nanjing University, Nanyang Technological University, Liblib AI · 2610.33106 · ▲ 40 · Deep dive
- Intrinsic Distribution Matching Inclusion AI, Westlake University, Tsinghua University · 2610.39633 · ▲ 37
- Read, Allocate, Route: Role-Coupled Tokenization for Pixel-Space Generative Transformers Peking University, Nanyang Technological University, SenseTime · 2610.40647 · ▲ 31
- WebAnything: From the Wild Web to Scalable Agent Environments Zhejiang University, Nanyang Technological University, Shanghai Qiji Zhifeng · 2610.37786 · ▲ 29
- WeVisCut: The Whole Editing Room in One Agent From Narration to Long-Form Video Xidian University, Tencent · 2610.38352 · ▲ 27
- Co-Design Bench: Multimodal Collaborative Coding Evaluation for Agentic Models Fudan University, Alibaba, Zhejiang University · 2610.35138 · ▲ 26
- IntReal: A Real-World Dataset and Benchmark for Multimodal Interleaved Generation Nanjing University, Meituan, Imperial College London · 2610.38425 · ▲ 24
- Multi-Token Prediction Done Right Massachusetts Institute of Technology · 2610.38331 · ▲ 22
- Online Dense Reinforcement Learning with Planning-Grounding Agent for UI Navigation on Live IOS Apps Apple · 2610.39264 · ▲ 19
- EditAgent: Rethinking Agentic Image Editing University of California San Diego, Adobe · 2610.39664 · ▲ 18
- AskStream: When and What to Ask Proactively under Uncertainty in Streaming Video Understanding Institute of Automation, Chinese Academy of Sciences, Beijing Academy of Artificial Intelligence, Hunan Normal University · 2610.38678 · ▲ 17 · Deep dive
- CoSTA: Compressed Speech Token Adaptation for Efficient Spoken Dialogue Models Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Wuhan AI Research · 2610.21143 · ▲ 15
- Youtu-Parsing-Omni: One Encoder, One Schema, All Modalities arXiv · 2610.39301 · ▲ 15
- Your LR Schedule May Not Be Principled: Effective Learning Rate Scheduling in Weight-Normalized Training ByteDance, Peking University · 2610.40362 · ▲ 15
- Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future? Prophetlab, RethinkLab, Shanghai Jiao Tong University · 2610.38342 · ▲ 14
- Glance: The Generator Is Its Own Reranker Airun Intelligence Labs · 2610.40454 · ▲ 14
- Training Foundation Models with a Single Algorithm Carnegie Mellon University, Johns Hopkins University, Aurora Innovation · 2610.40564 · ▲ 13
- After One-Click Video: The Inflow of Service-Type Content, the Failure of the Recommendation Pool Mechanism, and the BoLe System arXiv · 2610.37695 · ▲ 11 · Deep dive
- AB-Attention: Accurate Block Attention for Accelerating Video Diffusion Transformers InfRec, Chinese University of Hong Kong, Shenzhen, Shandong University · 2610.39663 · ▲ 11
- PhysInsert: Physically Plausible Object Insertion with Geometric Priors and Bidirectional Object-Scene Interactions Beihang University, MT Lab at Meitu, Tsinghua University · 2610.40453 · ▲ 11
- SO-101 Bench: Accessible Spatial Manipulation Independent researchers · 2610.36420 · ▲ 10
- OpenFlyV2: An onboard vision-language-action model for aerial perception and navigation Shanghai AI Laboratory, Northwestern Polytechnical University, Beihang University · 2610.39287 · ▲ 8 · Deep dive
- Broken on Arrival: Non-Functional Quantized LLM Artifacts in Public Model Registries and How to Catch Them Independent researchers · 2609.05881 · ▲ 7 · Deep dive
- Scaling Lean Theorem Proving for Real-World Code Verification Amazon, Case Western Reserve University · 2610.38174 · ▲ 7
- Matched Capabilities, Different Behaviors: A Controlled Study of MCP and CLI Interfaces for Language Agents Hong Kong Polytechnic University, Institute of Computing Technology of the Chinese Academy of Sciences · 2610.34298 · ▲ 6 · Deep dive
- Holo4D: Holistic 4D Reconstruction as Geometric Control for Video Diffusion University of Oxford, Nanyang Technological University · 2610.25276 · ▲ 6
- ABF-Bench: A Verifiable Benchmark for Agentic Blockchain Forensics Imperial College London, University of Oxford · 2610.23656 · ▲ 5
- Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models Massachusetts Institute of Technology, NVIDIA · 2610.38681 · ▲ 3 · Deep dive
- EPSBench: Benchmarking LLM Agents for Earnings Forecasting Georgia Institute of Technology · 2610.38293 · ▲ 3
- Unlocking Any-Order Generation in Pretrained Autoregressive Image Models Stony Brook University, International Laboratory on Learning Systems, LIVIA at ÉTS Montréal · 2610.39293 · ▲ 3
- FDSML: Formula-Driven Supervised Metric Learning with Parameter-Aware Triplet Loss Waseda University, Waseda Research Institute for Science and Engineering, National Institute of Advanced Industrial Science and Technology · 2610.39423 · ▲ 2
Thursday, October 29, 2026 · 17
- OmniShowcase: Omnidirectional Subject-Consistent Video Generation from Multi-View Images KAIST, Inha University, Fliption · 2610.35684 · ▲ 139
- Efficient Streaming Audio-Visual Target Speaker Extraction for Real-World Acoustic Scenes Qinghai University, Tsinghua University, Chinese Institute for Brain Research · 2610.32749 · ▲ 136 · Deep dive
- TAO: Tool-Aware Joint Scheduling for LLM-Based Agentic Serving Shanghai Jiao Tong University, Stanford University · 2610.37230 · ▲ 95
- Sensorimotor Alignment: A Platonic Proxy for Embodied Policy Success Rice University, NVIDIA · 2610.31670 · ▲ 56
- OmniSpectra: A Unified Foundation Model for Universal Spectra Representation Learning University of Virginia · 2601.15351 · ▲ 44 · Deep dive
- HuGe: Scaling Humanoid Generalist Vision-Language-Action Models with Egocentric Experience Westlake University, Zhejiang University, Shanghai Innovation Institute · 2610.34186 · ▲ 44
- RealCLI: Toward a Unified Evaluation of Human-Agent Collaboration for CLI Agents Beihang University, Renmin University, IQuest Research · 2610.22507 · ▲ 18
- PersonaCUABench: Benchmarking Computer-Use Agents Serving Diverse Personas University of Texas at Dallas, University of Southern California, Carnegie Mellon University · 2610.22990 · ▲ 17
- TauIndianBankBench: Procedure-Aware, Action-Oriented Evaluation of Tool-Using Conversational Agents National Payments Corporation of India · 2610.30161 · ▲ 11
- Amortizing Sequential Experimental Design in Reasoning LLMs University of Cambridge · 2610.20832 · ▲ 9
- LongZero: Enhancing Long-Context Reasoning in LLMs from Zero Data Tsinghua University · 2610.30132 · ▲ 8 · Deep dive
- Evaluating DeepSeek V4.1 Flash-1017 for loss of control and harmful manipulation risks Fangcun AI · 2610.35671 · ▲ 7
- LOB-Impact Benchmark: Emergent Market Impact in Generative Limit Order Book Models University of Oxford, University of California, Los Angeles · 2610.27100 · ▲ 4
- CausGround: Causal Interaction Grounding for Adaptive Embodied Planning East China Normal University, Beijing Academy of Artificial Intelligence, Northwestern Polytechnical University · 2610.29702 · ▲ 4
- PreView: Enhancing GUI Grounding with Preprocessed Views and Cues Independent researchers · 2610.33645 · ▲ 4
- Instruction Capacity: Behavior and Degradation of Language Models under Controlled Load Independent researchers · 2610.36095 · ▲ 4
- Know, then Draw: A New Learning Paradigm for Visual Generation University of California San Diego, Adobe · 2610.37626 · ▲ 2
Wednesday, October 28, 2026 · 51
- PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations Peking University, JD.com · 2610.37381 · ▲ 150 · Deep dive
- Hy-Nomas: Towards Free-form Interactive World Generation Tencent · 2610.36729 · ▲ 123
- Memory Tax: Dissecting the Net Value of Agent Memory University of Illinois Urbana-Champaign, Tsinghua University, Renmin University of China · 2610.34073 · ▲ 92
- HABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent Memory Shanghai Jiao Tong University, Shanghai AI Laboratory, Fudan University · 2610.24520 · ▲ 90
- Rollback: Experience Backtracking in World Models for Policy Improvement AgiBot, Beihang University, National University of Singapore · 2610.36068 · ▲ 84
- ChronoReplay: One Answer Interface Is Not Enough for Streaming Spatial Intelligence Westlake University, Zhejiang University, University of Science and Technology of China · 2610.34283 · ▲ 77
- ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning NVIDIA, KAIST · 2610.37666 · ▲ 62
- CodeCap-Bench: Can Code Serve as a Good Video Caption? Nanjing University, Kuaishou, TJU · 2610.37341 · ▲ 52
- LociArt: A Unified Paradigm for Free-Style Local Video Stylization Xiaomi, Xi'an University of Electronic Science and Technology · 2610.36456 · ▲ 48
- The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents InfRec Cardinal AI Lab · 2610.36077 · ▲ 44
- Can Discrete Tokens Serve as Semantically Lossless Representations of Visual Signals? Huazhong University of Science and Technology, Meituan LongCat, Shanghai Jiao Tong University · 2610.37662 · ▲ 38 · Deep dive
- RSI Arena: Towards Recursive Self-Improvement with Human Feedback Bake AI, University of Notre Dame, University of California, Los Angeles · 2610.33624 · ▲ 37
- On the SPOT: Enhancing Speculative Decoding for Diffusion Language Models via Commit-Order Prediction Technion, Israel Institute of Technology, Ben-Gurion University of the Negev · 2610.22543 · ▲ 33
- What Effort Level Should I Choose? On the Economics of Reasoning Effort in LLM Agents Johns Hopkins University, Together AI, Cornell University · 2610.34491 · ▲ 30 · Deep dive
- Can Test-Time Memory Evolution Outperform Test-Time Training for Scientific Discovery? UC Santa Barbara, University of California, Berkeley · 2610.32046 · ▲ 30
- Multiscale Policy Optimization for Long-Horizon LLM Agents Peking University, IQuest Research, Tsinghua University · 2610.35013 · ▲ 30
- VibeVideoBench: Evaluating Multimodal Agents on Code-Driven Video Generation Beijing University of Posts and Telecommunications, Zhongguancun Academy, Peking University · 2610.35864 · ▲ 30
- EmbeddingGemma 2: Unified and Efficient Multimodal Representations Google · 2610.37364 · ▲ 29 · Deep dive
- Zooming without Zooming v2: From Seeing Details to Connecting Evidence for Generalizable Multimodal Perception Shanghai Jiao Tong University, Ant Group, Zhongguancun Academy · 2610.37037 · ▲ 29
- Nemotron-Labs-Diffusion-Uni: Elastic Self-Refining Unified Diffusion LLM with Cross-Paradigm On-Policy Distillation NVIDIA · 2610.36466 · ▲ 28 · Deep dive
- Harness-Model Collaboration for Long-Horizon Continual Learning Huzhou Normal University, Alibaba, Zhejiang University · 2610.34724 · ▲ 21 · Deep dive
- Cross-modal Steering Vectors Exist Across Both Image and Text Generation Mila and McGill University, Carnegie Mellon University, Massachusetts Institute of Technology · 2610.36269 · ▲ 21
- MatAnyone-E: Sparse Memory Propagation for Efficient High-Resolution Video Matting University of Queensland, Nanyang Technological University, University of Adelaide · 2610.35724 · ▲ 19
- h1: A Taste of the Bitter Lesson in Automated Theorem Proving Meta, EPFL, ENPC, Institut Polytechnique de Paris · 2610.35962 · ▲ 19
- Channel Selection for Mixture-of-Experts Decoding Needs No Calibration Data KAIST, DeepAuto AI · 2610.37386 · ▲ 16 · Deep dive
- Omni-InteractScope: Benchmarking Proactive and Agentic Interaction in Continuous Omnimodal Streams Hong Kong Polytechnic University, Alibaba, Shanghai Jiao Tong University · 2610.35439 · ▲ 16
- Can LLMs Uncover Hidden Physical Laws through Free Exploration? Hong Kong University of Science and Technology, Guangzhou, Alibaba, Zhejiang University · 2610.37649 · ▲ 16
- MoveBench: A Rubric-Based Benchmark for Evaluating Large Language Models in Sports and Health Coaching Keep · 2610.26185 · ▲ 14 · Deep dive
- Timestral Mistral AI · 2610.37585 · ▲ 13 · Deep dive
- The Lexical Imprint of Pre-training Data NAVER Cloud, KAIST · 2602.10657 · ▲ 12
- Benchmarking Obstacle-Conditioned Safety in Household Robot Navigation Seoul National University · 2610.34674 · ▲ 11
- SearchHorizonBench: A Benchmark for Search Value in AI Agents Carnegie Mellon University · 2610.34714 · ▲ 11
- RynnWorld-Latent: A Cross-Embodiment World Model Conditioned on Universal Latent Actions Alibaba, Hong Kong Embodied AI Lab, Chinese University of Hong Kong · 2610.36791 · ▲ 11
- Your Policy Is Secretly a Harness: The Decision Geometry of Agent Interfaces Hong Kong University of Science and Technology, Guangzhou, Hong Kong University of Science and Technology · 2610.37255 · ▲ 8
- Rare, Relevant, and Still Wrong: How Memory Fails LLM Group Decisions University of Electronic Science and Technology of China, Xidian University, Nanjing University of Aeronautics and Astronautics · 2610.37624 · ▲ 8
- T-DIG: Text-Conditioned Diversity for Image Generation Carnegie Mellon University, Northeastern University · 2610.29003 · ▲ 7
- MLE-World: Predicting Machine Learning Experiment Outcomes with a Learned World Model University of Washington, University of Illinois Urbana-Champaign, University of Oxford · 2610.37175 · ▲ 7
- Truncate or Not? A Scaling Study of Looped Transformers with Truncated Backpropagation University of Hong Kong, Yale University · 2610.26030 · ▲ 6 · Deep dive
- OvisOCR3 Technical Report Alibaba · 2610.33066 · ▲ 6 · Deep dive
- Self-Assessment Reinforcement Learning for Search-State Awareness in Retrieval-Augmented Large Language Models Korea University, Amazon, AIGEN Sciences · 2610.30435 · ▲ 6
- What Outcome RL Buys: Finishing, Reliability, Rarely Reach Individual Researcher · 2610.34894 · ▲ 6
- Vision-Language-Action or World-Action? Just Action Models University of Tokyo, National Institute of Informatics · 2610.37219 · ▲ 6
- How Much Does the Hint Matter? Scope Conditions for Hindsight Self-Distillation in Financial Reasoning University of California, Berkeley, CTGT · 2610.36082 · ▲ 5 · Deep dive
- Learning and Reasoning with Geometry: How Generation Helps Spatial Understanding University of Hong Kong, Shanghai Jiao Tong University, ByteDance · 2610.36022 · ▲ 5
- Beyond the Estimate: Wiener-Informed Decoding for Easier Training of Pixel-Space Flow Models National University of Singapore · 2610.36192 · ▲ 5
- AlphaBet: Reinforcement Learning for LLM Event Forecasting Agents Georgia Institute of Technology, Amazon · 2610.22684 · ▲ 4
- Sub-monolayer tracking of molecular states at buried interfaces Washington State University, Australian Synchrotron, ANSTO, Prince Sattam bin Abdulaziz University · 2610.31924 · ▲ 4
- PhysEvolve: Verifier-grounded Self-evolution for Physical Reasoning with Executable Worlds Chinese University of Hong Kong · 2610.34921 · ▲ 4
- AdaPRA: Process Reward Agents with Adaptive Rubrics for Data Science Carnegie Mellon University, Microsoft · 2610.33245 · ▲ 3
- Quaternion Approximate Networks for Enhanced Image Classification and Oriented Object Detection Case Western Reserve University · 2509.05512 · ▲ 2
- Lipi-Ghor: A Large-Scale, Diarized, Multi-Domain Bengali Speech Corpus and a Scalable Pipeline for Building It Khulna University of Engineering and Technology, Bangladesh University of Engineering and Technology · 2610.36184 · ▲ 1
Tuesday, October 27, 2026 · 62
- Trajectory2Gym: Decision-Grounded Compositional Interaction Synthesis for Tool-Use Agents Westlake University, Ant Group · 2610.33568 · ▲ 368
- Trial by Trace: Benchmarking AI Agent Harnesses for Reliability and Failure Attribution Institute of Automation, Chinese Academy of Sciences, Zhongguancun Academy, Institute of Information Engineering, Chinese Academy of Sciences · 2610.32285 · ▲ 299
- StreamVibe: Proactive Video Agents for Evolving User Intent Google, University of California, Santa Barbara, University of California, Santa Cruz · 2610.34662 · ▲ 172 · Deep dive
- DynaFlow: Learning Free-Form Multi-Agent Orchestration National University of Singapore · 2610.34467 · ▲ 143 · Deep dive
- InteractiveWorld: A Verifiable Human-Centric Environment for Building Better Interactive MLLM Assistants Meta, Duke University · 2610.33966 · ▲ 127
- Work Agents: Systems, Capabilities, and Optimization Shanghai Jiao Tong University, University of Hong Kong, Hong Kong University of Science and Technology, Guangzhou · 2610.35845 · ▲ 119
- Representation Redistribution in Sparse Autoencoders University of Wisconsin-Madison · 2610.21240 · ▲ 62
- Continual Learning Exam: Benchmarking Long-Horizon RSI and Experiential Learning in LLM Agents arXiv · 2610.33956 · ▲ 61
- Spatial Audio Generation and Editing with Executable Plans and Self-Distilling Transfusion Hong Kong University of Science and Technology, Noiz AI, University of Washington · 2610.34415 · ▲ 55
- WorldCast: World-Consistent Character Video Generation with Large-Angle Camera Control Wuhan University, Xiaomi · 2610.33135 · ▲ 54 · Deep dive
- AV-FABLE: Audio-Visual Manipulation Forensics in Long-Form Videos Tsinghua University, Peking University · 2610.33413 · ▲ 48
- Eat Your Precision: Precision Objective Mitigates Reward Hacking in Rubric-based RL KAIST, Carnegie Mellon University · 2610.33512 · ▲ 48
- HiP3R: Hierarchical Pose Graphs for Long-Horizon 3D Reconstruction Nanyang Technological University, Imperial College London, ACE Robotics · 2610.31635 · ▲ 43
- Principia: Test-Time Hypotheses to Principle Construction for Scientific Discovery KAIST, DeepAuto AI · 2610.32195 · ▲ 42
- Qwen-Image-PE: Learning Visual Design for Text-to-Image Prompt Enhancement arXiv · 2610.33472 · ▲ 41
- World-in-Context: Interactive Video World Modeling with In-Context Conditioning Alibaba · 2610.33169 · ▲ 36 · Deep dive
- S3AC: Stabilized, Selective, and Scalable Actor-Critic for LLM Reasoning Shanghai AI Laboratory, Independent researchers · 2610.35518 · ▲ 33
- Evolving to Match Capabilities: Improving Harness Self-Evolution via Skill Adaptation University of Chinese Academy of Sciences, Independent researchers, University of California, Santa Cruz · 2610.33024 · ▲ 32
- Δτ-Bench: Pushing Agents Toward a Moving Capability Zhejiang University, University College London · 2610.34238 · ▲ 28 · Deep dive
- RynnVLA-Latent: Scaling Robotic Manipulation Foundation Models with 3D-Grounded Latent Actions Alibaba, Hong Kong Embodied AI Lab, Chinese University of Hong Kong · 2610.34356 · ▲ 28
- ProAS: A Progress-Aware Advantage Shaping Framework for Multi-Turn Agentic Reinforcement Learning Zhejiang University, Westlake University, Westlake Institute for Advanced Study · 2610.18683 · ▲ 27
- Trustworthy Recursive Self-Improvement in the Era of Large Models: Foundation, Risks, Evaluation and Open Challenges Renmin University of China, University of Chinese Academy of Sciences · 2610.35193 · ▲ 27
- Mycelium: A Meta-Study of Hallucination Detection Across Provenance, Typology, and Granularities Skoltech, Sber AI Lab · 2610.33608 · ▲ 26
- Build, Don't Play: Self-Improving LLM Research Agents Develop a Hearthstone Battlegrounds AI That Rivals Top Players Nanjing University, Shanghai AI Laboratory · 2610.33845 · ▲ 26
- miniTE: Diagnosing and Bridging the Execution Gap in Native NVFP4 Reinforcement Learning InfiX AI, NVIDIA · 2610.23928 · ▲ 24
- More Memory, Better Videos? A Controlled Study of Visual Memory in Long-Video Generation University of Science and Technology of China, Kuaishou, Tsinghua University · 2610.35001 · ▲ 23
- RoboHarness: A Simple Harness Outperforms VLA and World Action Models Zhejiang University · 2610.14018 · ▲ 22 · Deep dive
- How Far Can Coding Agents Go in Visual Generation? University of California, Los Angeles, Tsinghua University · 2610.19025 · ▲ 22
- CausalAlign: Causal Alignment of Video Generation Models for Consistent Ego-Motion Peking University, Beijing Institute for General Artificial Intelligence, NVIDIA · 2610.29915 · ▲ 22
- Extreme Compression of Per-Layer Embeddings via Cross-Layer Latent Vector Quantization Mohamed bin Zayed University of Artificial Intelligence, Westlake University, Zhejiang University · 2610.31309 · ▲ 21 · Deep dive
- iPrefBench: A User-Centric Benchmark for Personalized Mobile Use Agents Ohio State University, Apple · 2610.34735 · ▲ 20
- Thinking-in-Bets at Test Time Reduces Sycophantic Revisions in LLM Agents Google · 2610.34654 · ▲ 19
- EmbodiedLab: Benchmarking Embodied Agents That Act to Know HexaFuture, Tsinghua University, National University of Singapore · 2610.35385 · ▲ 19
- ReGE: Stronger Agents through Experience Synthesis University of Waterloo, University of North Carolina at Chapel Hill, University of British Columbia · 2610.33861 · ▲ 18
- ParetoRSI-Gym: Can AI Agents Improve Language Models Across Multiple Domains? University of Illinois Urbana-Champaign, University of California, Berkeley, Ohio State University · 2610.34386 · ▲ 16
- AgentProofBench: Benchmarking Reusable Verifiers for Web Agents Carnegie Mellon University · 2610.34634 · ▲ 16
- Steering Black-Box Large Language Model Agents with Pitfall Anticipation World Model University of Wisconsin-Madison, Amazon · 2610.29139 · ▲ 15
- Nested Sparsity for Scalable Pretraining and Continual Learning University of Hong Kong, University of California, Berkeley · 2610.31818 · ▲ 14 · Deep dive
- Rethinking Video Storage for AI Training: Breaking Cross-Frame Dependencies for Efficient Random Access Peking University, Tsinghua University · 2610.33079 · ▲ 14
- Novel or Not? PatEx-Gym: Agentic Prior-Art Search for Novelty Assessment in Patent Examination Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Dalian University of Technology · 2610.33503 · ▲ 13
- RoboJev: An Open Vision-Language Judge for Robot Success Detection Independent researchers · 2610.33718 · ▲ 13
- Beyond Benchmarks: Can Small Open-Weight Models Get Real Coding Work Done? Independent researchers · 2610.34241 · ▲ 13
- Joint or Per-Option? Compact Encoders for Closed-Set Decisions from 17 Million to 1 Billion Parameters Federal Institute of São Paulo · 2610.30970 · ▲ 12 · Deep dive
- CRM-Agent-Bench: Benchmarking Tool-Using Agents in Stateful Enterprise Environments Salesforce · 2610.32004 · ▲ 12
- DialFork: Going Beyond Factuality to Evaluate Procedural Hallucination in Conversational Agents arXiv · 2610.21561 · ▲ 11
- Skill^{infty}: Recursive Self-Improvement for Agent Skill Optimization National University of Singapore, City University of Hong Kong · 2610.35434 · ▲ 11
- Trust Mechanisms for Offline Generative AI at the Edge: A Connectivity-Regime Framework and an Empirical Study on a Raspberry Pi 5 Securagen AI · 2610.33732 · ▲ 9
- Rethinking Verification in Agent Harnesses Cornell University, Alibaba · 2610.34577 · ▲ 9
- Constrained Decoding: Harmful or Beneficial? Revisiting Its Impact on LLM Performance Télécom Paris, Institut Polytechnique de Paris, Nokia Bell Labs · 2610.35328 · ▲ 9
- OBB-PBD: Extending Parallel Box Decoding to Oriented Bounding Boxes for Referring Expression Grounding in Aerial Imagery Chungnam National University · 2610.35450 · ▲ 9
- Semantic Concept Steering Breaks the Explanation Drift Loop in Continual Learning Tel Aviv University, Open University · 2610.33877 · ▲ 8
- Intrinsic Goals for Autonomous Agents: Hierarchical Exploration Captures Future Volitional Action in Drosophila University of Maryland, Carnegie Mellon University · 2610.34638 · ▲ 7
- Assessing the CAD Capabilities of General-Purpose AI Models: A Case Study of GPT-6 Astra Zhejiang University · 2610.24451 · ▲ 6 · Deep dive
- OpenESM: Energy-Steered Models Are Scalable Foundation Learners Shanghai AI Laboratory · 2610.33005 · ▲ 6 · Deep dive
- EnvLunch: The Environment as a Free Lunch for Learning to Interact King's College London, Chongqing University, Alan Turing Institute · 2610.33817 · ▲ 6
- Did the Model Use That? Counterfactual Training Boosts Self-Attribution in Language Models Stanford University, Duke University, Emory University · 2610.35295 · ▲ 6
- Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations Case Western Reserve University · 2610.17117 · ▲ 5
- Learning from Automatic 3D Gaussian Splatting Edits for Online Bayesian Scene Change Detection University College London · 2610.28490 · ▲ 5
- ThinkBit: Search Codewords, Recalibrate Scales, Select by Reasoning Microsoft · 2610.34886 · ▲ 5
- Humans in Space: Can MLLMs Construct Human-Centric Spatial Understanding via Active Exocentric Exploration? University of Queensland, Adelaide University, University of Technology Sydney · 2610.34149 · ▲ 4
- Ψ-0.6: Adaptive Streams of Thought in a Unified Vision-Language World Model Stanford University · 2610.32177 · ▲ 3 · Deep dive
- CellTrack-Bench: Microscopic Cell Tracking Benchmark for Vision-Language Models Biohub · 2610.31566 · ▲ 2
Monday, October 26, 2026 · 45
- Reading, Writing, Sharing, and Interacting with the World through Video University of Oxford, National University of Singapore, Nanyang Technological University · 2610.31668 · ▲ 295 · Deep dive
- Function-Space Temporal Self-Distillation for Policy Optimization Indian Institute of Technology Kanpur · 2610.19687 · ▲ 140
- UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents Leiden University, Renmin University of China, Alibaba · 2610.30932 · ▲ 63
- HiSTEVE 1.0: Harnessing Human-Interface VLA Agents for Open-World Games University of Hong Kong, Zhejiang University, Fudan University · 2610.31162 · ▲ 57 · Deep dive
- RynnUnify: Towards Unified Embodied Brain Alibaba, Hupan Lab · 2610.32965 · ▲ 55
- RSIHub: A Composable Framework for Agent-Evolution Methods University of California, Berkeley, University of Science and Technology of China, ByteDance · 2610.32253 · ▲ 49
- On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models Google, Ulm University · 2610.32770 · ▲ 41
- Omni-Woodpecker: Hallucination Correction via Evidence-Prior Rebalancing for Omni LLMs Nanjing University, University of Chinese Academy of Sciences · 2610.27103 · ▲ 40
- Dual Exam: Controlled Evaluation of Multimodal Agents and World Models Hong Kong University of Science and Technology, National University of Singapore, Hong Kong Polytechnic University · 2610.31066 · ▲ 36
- DirCast: Live Esports Directing and Commentary with a Single Streaming Video-Language Model Alaya Lab, University of Tokyo, Georgia Institute of Technology · 2610.31734 · ▲ 32 · Deep dive
- DiscoDemo: Discovering Efficient and Diverse Robot Demonstrations for Imitation Learning KAIST, Holiday Robotics · 2610.32093 · ▲ 28
- Speculative Diffusion is a Tale of Two Tradeoffs Independent researchers, University of California, Los Angeles · 2610.31819 · ▲ 25
- Every Weight Is a Path of Shared Discrete Decisions ITMO University · 2610.31931 · ▲ 25
- TenetBench: A Clause-Anchored Russian Corpus for Evaluating Language Models against Professional Codes of Ethics Trusted AI Research Center of the Russian Academy of Sciences, Institute of Social Sciences · 2610.29477 · ▲ 20
- DraMaze: A Multi-Agent Pipeline with Future-Aware Memory for Diverse Microdrama Script Generation Fudan University, Shanghai Innovation Institute, Tencent · 2610.32002 · ▲ 20
- MMVistaReason: Toward Open-Data and Post-Training Recipes for Multimodal Reasoning Zhejiang University, Ant Group, Shanghai Jiao Tong University · 2610.01352 · ▲ 16
- SliderEdit++: Improved Slider Trajectories in Instruction-Based Image Editing Adobe, University of Maryland · 2610.14253 · ▲ 16
- Building Auditable AI Text Detectors with Sparse Autoencoders University of Pennsylvania, Google · 2610.30006 · ▲ 16
- FLOMSS: Generative Music Source Separation by Flow Matching Sony · 2610.17821 · ▲ 15
- Echo: Coupled Action and Task Status Signal Denoising for Vision Language Action Models Shanghai Jiao Tong University, Spirit AI, Tsinghua University · 2610.31569 · ▲ 15
- DuoLM: Fast and Effective Personalized Recommendation with Language Models Independent researchers, University of California, Irvine, Google · 2610.24481 · ▲ 13
- Generalized CycleGRPO for Fine-Grained Multimodal Understanding and Grounding National University of Singapore, University of Chinese Academy of Sciences, NTU · 2610.29552 · ▲ 12 · Deep dive
- LLM2Cert: LLM Synthesis of Certificates with SMT/Lean Verification Institute of Software, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Carnegie Mellon University · 2610.30524 · ▲ 12 · Deep dive
- Kolab: Benchmarking Human-Agent Collaboration in Long-Horizon Workflows University of Minnesota · 2610.21109 · ▲ 12
- MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation Alibaba, Zhejiang University · 2610.23234 · ▲ 12
- FocusVLA: Future-Intent Guided Dynamic View Routing for Vision-Language-Action Models Nanjing Institute of Software Technology, Institute of Software, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Nanjing · 2610.31423 · ▲ 12
- JockeyBench: Evaluating LLM Agents on Horse-Racing Decision-Making Hong Kong University of Science and Technology, Tsy Capital Limited, Independent researchers · 2610.19589 · ▲ 11
- Easy-3Dfy: Object-Level Geometric Composition for Single-Image 3D Scene Construction Zhejiang University, Tencent · 2610.32250 · ▲ 11
- Co-adaptive Rule Routing and Constraining for Structured Knowledge Retrieval University of Georgia, Michigan State University, Georgia Institute of Technology · 2610.14465 · ▲ 9
- PaperFormalizer: Author-in-the-Loop Autoformalization of Research Papers in Lean Stanford University · 2610.31810 · ▲ 9
- HoloBoxing: A Simple Recipe for Humanoid Boxing with 7 Minutes of Single-GPU Training arXiv · 2610.32909 · ▲ 9
- The fixed-point bundle method over product-of-simplex domains arising from game equilibria arXiv · 2609.16158 · ▲ 8
- DISC: Domain-Integrated Similarity Cost for Time-Series Dataset National Yang Ming Chiao Tung University, Dolby Laboratories, Adobe · 2610.31288 · ▲ 8
- RoboJudge: Multimodal Language Models as Judges for Embodied Video Generation Chinese University of Hong Kong, Westlake University, Northeastern University · 2610.27782 · ▲ 7
- Does LeWorldModel Preserve Local Action Geometry? Technological School Electronic Systems, Massachusetts Institute of Technology · 2610.31716 · ▲ 7
- KANLab: TensorFlow Playground for Kolmogorov-Arnold Networks University of Osnabrueck, University of Applied Sciences Osnabrueck · 2610.20117 · ▲ 6
- What Git Does Not Record: Provenance Legibility in Two AI-Assisted Research Repositories University of Texas at Austin, University of California, Los Angeles, University of Oxford · 2610.28655 · ▲ 6
- LoCA: Enabling Multi-Modality in Geospatial Location Encoders University of Athens, Cyprus Institute, Archimedes, Athena Research Center · 2610.32894 · ▲ 6
- A path-following framework on fiber bundle for variational inequalities arXiv · 2606.00778 · ▲ 5
- How Much Feedback Is Needed to Break Template Protection? The Geometric Vulnerability of Biometric Authentication Technion, Israel Institute of Technology, University of Haifa · 2610.25118 · ▲ 5
- Predictive Closure: A Set-Level Characterization of JEPA's Feature Selection Bar-Ilan University, University of California, Berkeley, Allen Institute for AI · 2610.27364 · ▲ 4
- Judging the Justification: Decision Certificate Supervision for Generative Reward Models University of Connecticut, NEC Labs America · 2610.27528 · ▲ 4
- WORTH Benchmark: Evidence Discovery under Not-Fully-Cooperative Information -- A Two-Episode Pilot Bread Studio · 2610.27597 · ▲ 3
- On the Token Collapse Phenomenon in Textual Graph Soft Prompting Sapienza University of Rome, Institute of Information Science and Technologies of the Italian National Research Council, University of Cambridge · 2610.31572 · ▲ 3
- FLORA: An Open Model Suite for Scaling Robotics Data Engines NVIDIA · 2610.31779 · ▲ 2
Friday, October 23, 2026 · 49
- WeVisOne-VL: Unified Visual Perception from Objects to Pixels Tencent · 2610.31329 · ▲ 253 · Deep dive
- Making MLLM Latents Matter: Factor-Aware Predictive Learning for Cinematic Video Generation Hong Kong University of Science and Technology, Tencent, École Polytechnique · 2610.30926 · ▲ 204
- Editing Images into Reusable Layers via Compositional Edit-Prior Distillation South China University of Technology, Tsinghua University, University of Science and Technology of China · 2610.24738 · ▲ 196
- Deep Search in Real-World Web Environments Hong Kong University of Science and Technology (Guangzhou), Fudan University, UW · 2610.30181 · ▲ 189
- IQuest-Q1: Advancing Foundation Capabilities for Agentic CLI Systems IQuest Research · 2610.30949 · ▲ 104 · Deep dive
- Toward Recursive Self-Improvement in Physical AI National University of Singapore, Huazhong University of Science and Technology, Shanghai Jiao Tong University · 2610.29677 · ▲ 76
- Deep Fusion Memory: A Chunk-Level Parametric Memory for Large Language Models Shanghai Jiao Tong University, Shanghai AI Laboratory · 2610.25100 · ▲ 60
- LaLa-World: Scaling Latent World Models with Natural Language Actions NAVER AI Lab, Georgia Institute of Technology, Brown University · 2610.27851 · ▲ 56
- Self-Distilling Future Video Representations for World Action Models KAIST, RLWRLD · 2610.30320 · ▲ 49
- Meta Expert: Cross-Layer Expert Sharing for Storage- and Retrieval-Efficient MoE Tsinghua University, Shanghai AI Laboratory · 2610.27598 · ▲ 47
- What Generation Knows: Unlocking Fine-Grained Understanding in Unified Multimodal Models Zhejiang University, University of Hong Kong, Chinese University of Hong Kong, Shenzhen · 2610.23987 · ▲ 45
- Code2Video-Compass: A Comprehensive Benchmark for General-Purpose Code-Based Video Generation across T2V, I2V, and V2V Fudan University, Sun Yat-sen University · 2610.28416 · ▲ 45
- Policy-Mask: Budget-Aware Adaptation of Diffusion Features to Discriminative Tasks HSE University, Moscow State University, Constructor University · 2610.28889 · ▲ 36
- Is Complex Agentic Memory Always Needed? Naive RAG Is Stronger Than We Thought Xi'an Jiaotong University, Shanghai Jiao Tong University, Zhejiang University · 2610.30755 · ▲ 36
- Heartbeat Thinking: Speculative Reasoning and Tool Execution While Listening in Realtime Voice Agents Ant Group · 2610.26233 · ▲ 34
- FilmFlow: Agentic System for 3D Virtual Film Cinematography Cornell University, Shanghai Jiao Tong University, University of California, Los Angeles · 2610.30753 · ▲ 34
- RLSplat: Resource-Linear Budget Control for Feed-forward 3D Gaussian Splatting KAIST, Korea University · 2610.27712 · ▲ 33
- SCOPE: Compiling Natural-Language Policies into Neurosymbolic Predicate Programs Plurai · 2610.31062 · ▲ 32
- Beyond Watching and Searching: Benchmarking Agents for Video-centered Deep Research Reports Peking University, Alibaba, Independent researchers · 2610.31158 · ▲ 32
- RouteWeaver: Weaving Mode Selection and Execution into Unified LLM Routing Nanyang Technological University, Huazhong University of Science and Technology, University of Illinois Urbana-Champaign · 2610.19705 · ▲ 31 · Deep dive
- Cola2: Co-Adaptive Hierarchical Language Modeling with Continuous Semantic Compression and Balanced Discrete Generation University of Hong Kong, ByteDance, Peking University · 2610.31305 · ▲ 28 · Deep dive
- Rethinking Sequence Compression with Test-time Regression Chinese University of Hong Kong, Independent researchers, Alibaba · 2610.31574 · ▲ 28
- Agent Mosaic: Recognizing Confidentiality Leaks in Agentic LLM Systems University of Southern California · 2610.30418 · ▲ 27
- AgentVQA: A Unified Benchmark for Agentic Visual Understanding University of Maryland, Princeton University · 2610.31494 · ▲ 25
- EvalResearchBench: Can AI Agents Build Their Own Feedback for Recursive Self-Improvement? Oregon State University, University of California San Diego, Pennsylvania State University · 2610.04184 · ▲ 24
- Verus-Insight: Towards Explainable Artifact Detection in AI-Generated Videos Peking University, Kuaishou, Hong Kong University of Science and Technology (Guangzhou) · 2610.30866 · ▲ 23
- Inherited, Not Invented: Tracing the Provenance of Wait in Reasoning Models Hong Kong University of Science and Technology, Peking University, Huawei · 2610.30882 · ▲ 23
- F2G-Pose: Geometry-Aware Foundation Feature Lifting for Direct RGB-D Category-Level Object Pose Estimation Fudan University, Shanghai Innovation Institute · 2610.31623 · ▲ 22
- Diffusion Language Models Change Speed Across Tasks and Languages University of California, Los Angeles, LMU Munich, Munich Center for Machine Learning · 2610.25835 · ▲ 21 · Deep dive
- Different Tasks, Different Timesteps: Temporal Task-Aware Quantization for Diffusion Language Models Technion, Israel Institute of Technology, Ben-Gurion University of the Negev · 2610.31198 · ▲ 21 · Deep dive
- RTAformer: Relation-Transition-Aware Transformer for Video Scene Graph Generation Sejong University · 2610.20091 · ▲ 21
- AnyGameRenderer: Entity-Consistent Generative Rendering Across Games and an Entity-Level Benchmark Alaya Lab · 2610.31466 · ▲ 21
- Foundation4D: Towards a unified perception and reconstruction model for autonomous vehicles NVIDIA · 2610.30340 · ▲ 20 · Deep dive
- MolLean: A Process-Verifiable Language for Chemical Reasoning National University of Singapore · 2610.20459 · ▲ 19
- SEAM: Spectral Embedding Alignment and Merging for Low-Resource TTS ITMO University, St. Petersburg Federal Research Center of the Russian Academy of Sciences, MWS AI · 2610.26252 · ▲ 17 · Deep dive
- RIDAR: Safeguarding Research Integrity in Autonomous Research Agents via Automated Detection and Runtime Intervention Tsinghua University, Renmin University of China, Zhipu AI · 2610.31372 · ▲ 17
- BrickWright: Agentic Creation of Physically Feasible Brick Builds Carnegie Mellon University, Independent researchers, Stanford University · 2610.31651 · ▲ 16
- MambaAct: Efficient Mamba Action Experts for Flow-Matching Vision-Language-Action Models Mohamed bin Zayed University of Artificial Intelligence, Technical University of Denmark · 2610.22951 · ▲ 15 · Deep dive
- SplatLab: Autonomous Scientific Discovery for 3DGS Research University of California San Diego · 2610.30468 · ▲ 13
- What Makes a Good Unmasking Policy? A Controlled Study of Learned Unmasking in Diffusion Language Models Nanyang Technological University, Singapore, Institute for Advanced Intelligence and Computing at A-STAR, Centre for Frontier AI Research at A-STAR · 2610.31576 · ▲ 13
- The LLM-Attacker: A Unified Red-Team Framework for Privacy Evaluation of Split and Distributed LLM Systems Setloop.io · 2610.22722 · ▲ 12
- Skeleton-Grounded Latent Action Model for Robot World Models Harbin Institute of Technology, Shenzhen, National University of Singapore, Ping An Technology · 2610.31474 · ▲ 12
- TILDE: Teacher-in-the-Loop Distillation from Embodied Foundation Models University of Hong Kong · 2610.25376 · ▲ 11
- HyperDrone: Simulation-to-Reality Transfer at Millions of Frames per Second University of California, Berkeley · 2610.28950 · ▲ 11
- IF-MAP: Influence-Function-Inspired Agent Prioritization in Multi-Agent Systems Chinese University of Hong Kong, Shenzhen, Harbin Institute of Technology, Harvard University · 2610.24383 · ▲ 10 · Deep dive
- Realism VS Accuracy: Event Sequence Forecasting from a Generative Modeling Perspective Applied AI Institute, ITMO University, Vega Institute Foundation · 2610.12586 · ▲ 10
- Who Takes the Lead? Model Ability and Task-Conditioned Collaboration with Humans McGill University, Massachusetts Institute of Technology, University of Cambridge · 2610.27148 · ▲ 10
- Bridging Exploration and Persona: Information-Directed Dueling for Test-Time Agent Personalization Chinese University of Hong Kong, Shenzhen, University of Copenhagen, City University of Hong Kong · 2610.24385 · ▲ 9 · Deep dive
- TARSphere: A Robot-Oriented Full-Stack Framework for Long-Horizon Embodied Agents in Evolving Physical and Social Environments Zhejiang University, Fudan University · 2610.30043 · ▲ 8
Thursday, October 22, 2026 · 58
- Spectral-Spatial Interpretation University of Wisconsin-Madison · 2610.21114 · ▲ 732
- StreamEval: Evaluating Interactive Streaming Video Generation Models Tsinghua University, NVIDIA, Carnegie Mellon University · 2610.28402 · ▲ 582
- Right Screen, Wrong Transition: World Models as Verifiers for GUI Agents Nanyang Technological University, Institute of Science Tokyo, Sony · 2610.11942 · ▲ 188
- Looking Ahead: Modeling Future Evolution in Autoregressive Video Generation Nanjing University, Tencent, Zhejiang University · 2610.29989 · ▲ 172
- SpatialScout: Can VLM Agents Actively Explore and Map 3D Scenes? Peking University, Tianjin University · 2610.26659 · ▲ 170
- DONIC: Generalist Dexterous Bimanual Tracking Control from Human Demonstrations Ropedia · 2610.29997 · ▲ 161
- Cache Engineering for Agents Tsinghua University, Chinese University of Hong Kong · 2610.28577 · ▲ 159
- Agent World Model in the Wild: Computer-Use Agent Environment Synthesis University of North Carolina at Chapel Hill, Microsoft, Georgia Institute of Technology · 2610.26828 · ▲ 101
- Scaling in Games: Continued Pre-Training For Embodied Agents In Diverse Virtual Worlds miHoYo, Tsinghua University, University of Chinese Academy of Sciences · 2610.28698 · ▲ 78
- PrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric Synthesis University of Hong Kong, Nanjing University, Fudan University · 2610.20538 · ▲ 72
- Retrieval Augmented Generation Compositionality IBM, Weizmann Institute of Science · 2610.29722 · ▲ 54
- TraceForge: Distilling with Your Own Thinking Independent researchers · 2610.29680 · ▲ 50
- Socratic Skill Learning: Interleaving Teaching and Practice for Transferable Agent Skills Nanjing University, Alibaba · 2610.17860 · ▲ 49
- AgenticRoboRL: Bringing Coding-Agent Post-Training to Agentic Robotics University of Central Florida, University of Southern California, Michigan State University · 2610.28589 · ▲ 49
- Building AI Research Teams for Autonomous Discovery of Linear Attention Architectures UC Santa Barbara, Pennsylvania State University, LinkedIn · 2610.28887 · ▲ 45
- Memory-Centric Continual Learning for AI Agents arXiv · 2610.29357 · ▲ 43
- CulturalRiddles: A Multicultural Benchmark of Riddles Cohere Labs Community, Lida Safety, Aarhus University · 2610.29968 · ▲ 43
- Think Harder Where It Matters: Item-Level Adaptive Recursion for Sequential Recommendation Renmin University of China · 2610.30128 · ▲ 41
- Qwen-Audio-3.1-ASR Technical Report Alibaba · 2610.28711 · ▲ 37
- CompactSplat: Spatially Adaptive Gaussian Distribution for Feedforward Scene Reconstruction Shanghai Jiao Tong University · 2610.13177 · ▲ 34
- ALPS: Visual State Encoders for World Models University of California, Santa Cruz, Google, University of California, Santa Barbara · 2610.30120 · ▲ 33
- DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agents Google, Boston University, University of Washington · 2610.19376 · ▲ 32
- From VLM Planners to Robotic Agents: Self-Evolving Harnesses for History-Dependent Manipulation University of Science and Technology of China, Shanghai AI Laboratory, Shanghai Jiao Tong University · 2610.29560 · ▲ 32
- VerifierBench: Benchmarking the Generation of Executable Verifiers for LLM Agents University of Pennsylvania, University of Illinois Urbana-Champaign, University of California, Berkeley · 2610.28679 · ▲ 31
- Automating ML for Science: Can Frontier Agents Climb Scientific Hills in the Wild? Anthropic Fellows Program, Anthropic · 2610.28763 · ▲ 29
- Harness Scaling for Generalizable Agentic Reinforcement Learning Renmin University of China, IQuest Research, Beihang University · 2610.29458 · ▲ 28
- RealMe: Immersive Volumetric Video from Minimal Captures Zhejiang University, Four DV.ai · 2610.30023 · ▲ 28
- Real In, Real Out: What If We Only Use Real Data for Scene Text Editing? Fudan University, Shanghai Key Laboratory of Multimodal Embodied AI, Tencent · 2610.27592 · ▲ 27
- Agentic World Action Models University of California, Berkeley, NVIDIA, University of Pennsylvania · 2610.30136 · ▲ 25
- Sparsity Should Be ALLOTted by Output: Training-Free Allocation of Block-Sparse Attention for Video Diffusion Transformers University of Chinese Academy of Sciences, Institute of Automation, Chinese Academy of Sciences, Peking University · 2610.29441 · ▲ 24
- PACE: Periodic and Channel-Aligned Encoding of Four-Bit Recurrent States Humboldt University of Berlin · 2610.29545 · ▲ 24
- Multimodal Embodied Agents: A Unified Capability-centric Perspective from Computer-Use to Robot-Use National University of Singapore · 2610.30108 · ▲ 24
- Entertain, then Cancel: How Thinking Makes Reasoning Models Read Some and Or Literally Hong Kong University of Science and Technology, Hong Kong University of Science and Technology, Guangzhou, Peking University · 2610.21204 · ▲ 23
- Language Models Learn to Improve Language Models Pennsylvania State University, University of California, Santa Cruz, Emory University · 2610.23344 · ▲ 22
- Which Agent's Skill Should Change? Group-Relative Credit Assignment for Evolving Multi-Agent Teams Mayo Clinic, University of Manchester, Alibaba · 2610.27227 · ▲ 22
- WorldStream: Streaming Interactive 3D Worlds with Hierarchical Planning and Persistent State University of Science and Technology of China, JD.com · 2610.28827 · ▲ 20
- SpatialBrain: Spatial Intelligence Growing from Experience Shanghai Jiao Tong University, Zhejiang University · 2610.29317 · ▲ 20
- Evolving with Open Worlds: Towards Recursive Self-Improvement for Future Forecasting Renmin University of China, IQuest Research, Nanyang Technological University · 2610.29784 · ▲ 19
- Hidden Messages Between Diffusion Language Model Agents: Causal Evidence for Latent Communication ITMO University, HSE University · 2610.28905 · ▲ 18
- ACE-Brain-0.6: Scaling Spatial Grounding for Physical Foundation Models arXiv · 2610.27040 · ▲ 17
- StyleStream 2.0: Fast and Controllable Streaming Voice Style Conversion University of California, Berkeley · 2610.28940 · ▲ 17
- Toward Part-Aware Choral Transcription with singing voice assignment UNSW Sydney, University of Western Australia, Queen Mary University of London · 2610.09861 · ▲ 16
- Answering in the User's Variety: Calibrated Accommodation Pairs for Post-Training Hong Kong University of Science and Technology, Hong Kong University of Science and Technology, Guangzhou, Peking University · 2610.28474 · ▲ 16
- ToM2Vec: Continuous Theory-of-Mind Tokens for Multi-Agent Coordination University of North Carolina at Chapel Hill, Cisco, University of Texas at Austin · 2610.29934 · ▲ 16
- RubatoCodec: Budgeted Adaptive Speech Tokenization with Streaming Decoding Third Intelligence · 2610.29434 · ▲ 15
- From RL Papers to Framework Code: Inclusion and Review in Seven Open-Source Training Frameworks University of Melbourne · 2610.28027 · ▲ 14
- DiffResearch: Accelerating Deep Research with Diffusion Language Models ITMO University, AXXX · 2610.28909 · ▲ 13
- Quant-Arena: Benchmarking LLM Agents for Quantitative Trading Research Navers Lab at Einsia AI · 2610.30113 · ▲ 13
- VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage - An empirical study of unknown-attack honesty, cascade gray zones, and calibration under drift Independent researchers · 2610.30090 · ▲ 12
- Looped Uniform Diffusion Language Models Mohamed bin Zayed University of Artificial Intelligence, Sony, Massachusetts Institute of Technology · 2610.30066 · ▲ 11
- Design Anything, Design at Scale Tsinghua University, Beijing Institute of Fashion Technology, Massachusetts Institute of Technology · 2610.24771 · ▲ 6
- Stratified Sparse Attention: Reducing the Sparse-to-Dense Gap at Extreme Sparsity Rice University, University of California, Berkeley, Stanford University · 2610.28441 · ▲ 6
- Exploring The Empirical Privacy Landscape of Synthetic Data Johns Hopkins University, OpenAI · 2610.22659 · ▲ 5
- Bridged Self-Distillation for On-Policy Learning Technical University of Munich, LMU Munich, Munich Center for Machine Learning · 2610.26397 · ▲ 4
- MMedAgent-Plus: Scaling Multimodal Medical Agents across Modalities, Tasks and Workflows Duke University, Stanford University, Yale University · 2610.27233 · ▲ 4
- FrontierCS 2: A Benchmark for Computer Science R&D University of California, Berkeley, Princeton University, University of Waterloo · 2610.29932 · ▲ 4
- Same Simulation, Different Evidence: Where Retrieved Tool Output Costs LLM Agents Accuracy University of California, Riverside, University of California, Irvine · 2610.26836 · ▲ 3
- RoboCasa-Shift: Stage-Aware Robustness Evaluation of Vision-Language-Action Policies in Long-Horizon Household Manipulation Fudan University, Chinese University of Hong Kong, Shenzhen Loop Area Institute · 2610.27861 · ▲ 3
Wednesday, October 21, 2026 · 64
- Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds Skywork AI, Chinese University of Hong Kong, University of California San Diego · 2610.27999 · ▲ 194
- PAR: Progress-Aligned Recurrence for Length Generalization in State Tracking Ant Group, National University of Singapore · 2610.27628 · ▲ 162
- Ropedia W0: Scaling Loco-Manipulation World Models across Embodiments with Human Experience Nanyang Technological University, Ropedia, Duke University · 2610.26855 · ▲ 151
- CoEvo-RSI: When Harnesses Evolve, Which Models Should Follow? National University of Singapore, Agency for Science, Technology and Research, SenseTime · 2610.25078 · ▲ 120
- S-Agent 2.0: Spatial Reasoning with a General Coding Agent Nanyang Technological University, University of Hong Kong, Ropedia · 2610.28354 · ▲ 106
- DecisionPivot: A Diagnostic Benchmark for Decision Models in LLM Systems StartLux · 2610.28522 · ▲ 74
- Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition Shanghai AI Laboratory · 2610.28578 · ▲ 72
- VLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action Reuse Zhejiang University · 2610.25043 · ▲ 64
- Learning the 4D World through Point Trajectory Forecasting KAIST, ETH Zurich, Seoul National University · 2610.28594 · ▲ 52
- SolarRenderer: Few-Step Generative Reference Rendering National University of Singapore, Lightspeed, Chinese University of Hong Kong, Shenzhen · 2610.28630 · ▲ 52
- Correction Training for Diffusion Models University of California, Berkeley, Reve · 2610.14136 · ▲ 48
- Focused Omni-Modal GUI Agents Nanyang Technological University, Alibaba · 2610.26203 · ▲ 46
- PhysSkin++: Real-Time and Generalizable Physics-Based Animation via Self-Supervised Neural Skinning Zhejiang University, Ant Group, Beijing Institute for General Artificial Intelligence · 2610.26067 · ▲ 37
- MMEnterprise-Bench: Evaluating Enterprise Agents over Evolving Multimodal Records KAIST, Korea University · 2610.27911 · ▲ 33
- SWIFT: Self-guided Weak-to-strong dIFfusion models via Target extrapolation University of Michigan, Ann Arbor, Alibaba, University of California, Berkeley · 2610.25844 · ▲ 32
- SpecKCast: Spectral Kalman Forecasting for Diffusion Acceleration University of Toronto, Vector Institute, Samsung · 2610.26837 · ▲ 28
- EvaOmni: An Evidence-Grounded Benchmark for Audio-Visual Reasoning Shanghai Jiao Tong University, Shanghai Innovation Institute, Inner Mongolia University of Technology · 2610.28444 · ▲ 27
- Joints, Camera, Action! Co-Generating 3D Motion and Video for Arbitrary Rigs NVIDIA, Technion, Israel Institute of Technology, Bar-Ilan University · 2610.28632 · ▲ 26
- Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative Offloading Shanghai Jiao Tong University · 2511.14102 · ▲ 25
- LLM2R: From Large Language Models to 3D Reassemblers National University of Singapore, Hong Kong Polytechnic University · 2610.18604 · ▲ 25
- FlowState: Execution State as Memory for Long-Horizon LLM Agents Ant Group · 2609.34565 · ▲ 23
- Epi: Towards a Persistent, Self-Improving Agent for Autonomous Research Bake AI, University of Notre Dame, University of Washington · 2610.24944 · ▲ 23
- Breaking Diffusion: Parallel Reasoning with Draft Refinement KAIST, Microsoft, SkillBench · 2610.26542 · ▲ 23
- Ask to Distill: Learning to Synthesize Data for On-Policy Distillation University of Illinois Urbana-Champaign, Hong Kong University of Science and Technology, Texas A and M University · 2610.27450 · ▲ 23
- Evolving Agent Interfaces University of California, Santa Barbara, Cornell University, Carnegie Mellon University · 2610.28601 · ▲ 23
- Large Flow Language Models University of California, Berkeley, Amazon, KAIST · 2610.28559 · ▲ 22
- EAU: Environment-Agent-User Co-evolution for Training User-Centric Agents University of Alberta, National University of Singapore, Kuaishou · 2610.16523 · ▲ 21
- Painting Change Map as guidance for generating painting processes University of Trento · 2610.25369 · ▲ 20
- Training the Whole Agent, Not Just the Model Zhejiang University, Lightspeed, Independent researchers · 2610.28069 · ▲ 19
- Event-STVG: Spatio-Temporal Video Grounding for Long-Form Event Reasoning Tsinghua University, Beijing University of Posts and Telecommunications · 2610.22444 · ▲ 18
- From Short Clips to Real-Time Streams: Identity-Aware Supervision and Context-Efficient Generation for Video Face Swapping Zhejiang University, Zhejiang University of Technology · 2610.26681 · ▲ 18
- Muse Glimmer 30B: An On-Device Agentic Model Meta · 2610.24628 · ▲ 16
- EdgeKernelBench: Unified LLM-Driven Kernel Generation for Edge Backends Westlake University, Zhejiang University · 2610.24648 · ▲ 16
- More Memory, Worse Agents: Error Reproduction and Anti-Persistence in LLM Agents Indiana University, Tencent · 2610.27406 · ▲ 16
- Good Judgment, Bad Advice: Decision-Model Controllers Raise Agent Cost Without Improving Success arXiv · 2610.24649 · ▲ 15
- HarnessRL: Co-Evolving Policies with a Just-in-Time Training Harness for Open-Ended RL Alibaba, University of Montreal, Mila · 2610.26426 · ▲ 15
- HybridResearcher: Bridging the Structured-Unstructured Boundary for Deep Research Agents Seoul National University, University of California, Los Angeles, University of North Carolina at Chapel Hill · 2610.24899 · ▲ 13
- Unsupervised Tool Evolution for Coding Agents in Data-Lake Analytics University of Southern California, Google, New York University · 2610.27327 · ▲ 13
- On the Limits of Scale: Inverse Scaling in Vision-Language Models University of Edinburgh, Imperial College London, Miniml AI · 2610.27360 · ▲ 13
- Harness Self-Evolution for Long-Horizon Terminal Agents Tencent, Indiana University, University of Maryland · 2610.27435 · ▲ 13
- Evaluating Agentic Capabilities Before Post-Training: An End-to-End Approach Tencent, Shanghai Jiao Tong University · 2610.19425 · ▲ 11
- On the Fragility of Tasks under Composition: Implicit Budgeting in Language Models Technical University of Munich, Massachusetts Institute of Technology · 2610.27190 · ▲ 11
- Self-Credit: Credit Assignment for Long-Horizon Agentic RL via Intra-group Rescaling Meta, New York University, Duke University · 2610.28388 · ▲ 11
- PhysLDM: Latent Diffusion for High-Fidelity Deformable Simulation Nanyang Technological University, Shanghai AI Laboratory · 2610.07609 · ▲ 10
- PVBench: Quantifying Physical Plausibility in Video Generation with Object-Level Physical Measurements Zhejiang University, Huazhong University of Science and Technology, Tencent · 2610.26711 · ▲ 10
- Beyond Compilation: Evaluating Faithful Natural-Language-to-Lean Statement Formalization University of California, Riverside, University of Arizona, University of California San Diego · 2606.31002 · ▲ 9
- WinARDet: Compressed Window Auto-Regression for Long-Horizon Vision-Language Grounding Nankai University, Beijing Normal University · 2610.19419 · ▲ 8
- Multimodal Reranking Is a System One Decision Aalto University · 2610.28435 · ▲ 8
- λ_0: Scaling Open-World Egocentric Human Data for Humanoid Loco-Manipulation Alibaba · 2610.18950 · ▲ 7
- On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs Seoul National University · 2610.23400 · ▲ 7
- EMBER: Predicting TCR--pMHC Binding Without a Docked Complex Northwestern University · 2610.27334 · ▲ 7
- OverSeer: Multi-Agent World Models Beyond What Agents See Meta, University of Oxford · 2610.28554 · ▲ 6
- OmniCam: Omni-Camera Trajectory Generation via Geometry-Grounded Pose Token Learning Fudan University, Shanghai Innovation Institute, Tencent · 2610.09513 · ▲ 5
- ATEST: Adaptive Resource-Aware Test-Time Scaling for LLM Inference Iowa State University · 2610.25519 · ▲ 5
- Slot-JEPA: Object-Structured Prediction Interfaces for Latent World Models University of Chinese Academy of Sciences, Institute of Computing Technology, Chinese Academy of Sciences, Beijing Academy of Artificial Intelligence · 2610.28074 · ▲ 5
- HRoPE: Homography Relative Positional Encoding Zhejiang University, Westlake University, Huazhong University of Science and Technology · 2610.28284 · ▲ 5
- HumanSkill-4D: Training-Free Language Querying of Animatable Gaussian Avatars in Space and Time Ulsan National Institute of Science and Technology, NAVER AI Lab · 2610.24452 · ▲ 4
- TRACE: Tracking Multimodal Clinical Trajectories for Inpatient Risk Prediction Yeji X · 2610.25765 · ▲ 4
- Axiomatic World Modeling for Physics Reasoning Northwestern University, University of Florida · 2610.24519 · ▲ 3
- VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Token Korea University, Adobe, KAIST · 2610.26857 · ▲ 3
- What Makes a Good Symbol for Hybrid Discrete-Continuous Planning? Carnegie Mellon University, University of Pennsylvania · 2610.23010 · ▲ 2
- World Models with Configurable Perception East China Normal University, Hong Kong University of Science and Technology · 2610.27914 · ▲ 2
- Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models Case Western Reserve University · 2603.19233 · ▲ 1
- BiDex-100: A Comprehensive Benchmark for Bimanual Dexterous Manipulation Across 100+ Simulation Tasks Zhejiang University, Ant Group · 2610.25459 · ▲ 1
Tuesday, October 20, 2026 · 75
- Scaling Parameter-Efficient Fine-Tuning for Brain Encoding Columbia University, Mind Lab, Tsinghua University · 2610.26802 · ▲ 246
- EgoXperience: Learning World-Grounded Egocentric Representations from Human Experience Ropedia, Nanyang Technological University, Tsinghua University · 2610.26625 · ▲ 188
- Holi-World: Building Holistic World Intelligence from Human-Ego Experience Ropedia, Nanyang Technological University, Hong Kong University of Science and Technology · 2610.26549 · ▲ 172
- MemT-1.0: Towards Memory-Native Evolution MemTensor · 2610.26825 · ▲ 135
- Learning from Fleet Experience: A Collective Learning Paradigm for Robots Tsinghua University, AgiBot, Infinigence AI · 2610.25736 · ▲ 127
- OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory University of California, Los Angeles, Apple · 2610.26860 · ▲ 114
- ScaleNav: Scalable Internet Video Pretraining for Language-Conditioned Local Navigation Zhejiang University, Li Auto, Shenzhen University · 2610.25690 · ▲ 91
- SpecMRL: Redesign Matryoshka Representation Learning within Parametric Spectrum Renmin University of China, Lenovo Group, Wuhan University · 2610.26824 · ▲ 87
- Self-Evolving Coding Agents via Harness-Policy Optimization Microsoft, University of Virginia, Carnegie Mellon University · 2610.22791 · ▲ 55
- CoworkEnv: Synthesizing Training Environments for Long-Horizon Cowork Agents Hong Kong University of Science and Technology, Chinese University of Hong Kong, Hong Kong Baptist University · 2610.24960 · ▲ 51
- Interlock: From Local Correctness to Task Completion in Long-Horizon Agents Peking University, Hong Kong University of Science and Technology, Guangzhou, Tsinghua University · 2610.19686 · ▲ 48
- OR-License: Evaluating LLM Agents as Real-World Operations Research Practitioners Hong Kong University of Science and Technology, Guangzhou, University of Hong Kong, Qingzhi Optimization · 2610.24087 · ▲ 48
- AnyGoal: On World Model Generalization Beyond Expert Rollouts Brown University, Cold Spring Harbor Laboratory, Mila and the University of Montreal · 2610.26844 · ▲ 45
- MT Forcing: Towards Stable and Consistent Multi-Turn Image Editing with Autoregressive Diffusion SenseTime, Hong Kong University of Science and Technology, Stanford University · 2610.25404 · ▲ 44
- RoAR-VAE: Improving Latent Diffusability through Robust Prediction Tsinghua University, Microsoft · 2610.26216 · ▲ 39
- Kronecker Delta Rule for Long Context Linear Attention Johannes Kepler University Linz, NXAI · 2610.26834 · ▲ 38
- BlenderWorld: Benchmarking Visual Agents for 3D Creation from Real-World References Zhejiang University, Alibaba, Chinese University of Hong Kong · 2610.23986 · ▲ 37
- MemPilot: Orchestrating On-Demand Multimodal Memory Curation for LLM Agents Nanyang Technological University, Tsinghua University, University of Illinois Urbana-Champaign · 2610.06830 · ▲ 36
- InfiniDream: Open-Ended Photorealistic 3D Worlds Through Procedural Generation Alaya Lab, National Yang Ming Chiao Tung University, National Taiwan University · 2610.26853 · ▲ 34
- AdaExploreBench: Benchmarking LLM Agents for Proactive Exploration and Test-time Adaptation in Evolving Environments Hong Kong University of Science and Technology, Hong Kong University of Science and Technology, Guangzhou, HKUST UGAIL lab · 2610.25950 · ▲ 33
- Manga2Animate: An Agentic Framework for Faithful Manga Animation with Script and Storyboard Planning National Yang Ming Chiao Tung University · 2610.26851 · ▲ 33
- HarnessPro: Recursive Harness Self-Improvement with Pareto Optimization for Performance-Cost Trade-offs Tsinghua University, Z.AI, Shanghai Jiao Tong University · 2610.26770 · ▲ 32
- FineSplat: Fine Patches, Local Groups, Detailed Splats Yonsei University · 2610.25011 · ▲ 30
- AffordanceWAM: Affordance-Aware Joint World-Action Modeling for Robot Manipulation University of Hong Kong, Peking University, Hong Kong University of Science and Technology, Guangzhou · 2609.22332 · ▲ 24
- Think Visually Across Space and Time: Grounded Reasoning for Video Segmentation Harbin Institute of Technology, Shenzhen, Pengcheng Laboratory, Pazhou Lab · 2610.25090 · ▲ 24
- Evidence-First Agentic Verification Scale AI, University of Arizona · 2610.17724 · ▲ 22
- Frechet Forcing: Training Long-Horizon Streaming World Models from Short Clips arXiv · 2610.25666 · ▲ 22
- Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy Galbot · 2610.26848 · ▲ 22
- LIBERO-Industry: Benchmarking Robotic Manipulation under Industrial Process Constraints Tongji University, Chinese University of Hong Kong, University of Aberdeen · 2610.16782 · ▲ 21
- RoboValue: A Fine-Grained Sim-and-Real Benchmark for Unified Evaluation of Robotic Value Models Tsinghua University, AgiBot, Xspark AI · 2610.26663 · ▲ 21
- Harnessing Textual Refusal Directions for Multimodal Safety University of Trento · 2606.31876 · ▲ 20
- DeOverlay: A Benchmark and Source-Anchored Flow for Video Overlay Removal ByteDance, Southern University of Science and Technology · 2610.26081 · ▲ 19
- ECHO: Efficient Causal Self-Recursion Beyond Auxiliary Multi-Token Prediction Baidu, Independent researchers · 2610.24766 · ▲ 17
- State World Models: Execute the Rules, Generate the Pixels Tsinghua University, GigaAI · 2610.26811 · ▲ 17
- Self-Evaluating Recursive Agents Tongji University, Independent researchers, Shanghai AI Laboratory · 2610.04902 · ▲ 16
- ACEdit: Stable Multi-Turn Image Editing via Asymmetric Conditioning Tel Aviv University · 2610.22328 · ▲ 16
- Progress-Aware Distillation for Mitigating Stagnation in Small Language Model Agents Pennsylvania State University, Microsoft · 2610.19439 · ▲ 15
- Factorization-Guided Training and Inference for Diffusion-Drafted Speculative Decoding Virginia Tech, Carnegie Mellon University, ByteDance · 2610.23935 · ▲ 15
- Policy Is Not Capability: Some Safety Benchmarks Measure Which Definition of "Unsafe" You Trained On Lenovo · 2610.25060 · ▲ 15
- Empowering Edge Agents: A Systematic Analysis of Edge-Cloud Collaboration in LLM Agent Execution Northeastern University in Shenyang, Tsinghua University, China Mobile Online Services · 2610.17816 · ▲ 14
- Co-Evolving Acting and Forecasting for Language Agents University of Illinois Urbana-Champaign, University of California San Diego, Meta · 2610.20597 · ▲ 14
- Learning to Discriminate Scene Structures Makes Self-Supervised Depth Learning Scalable Tongji University · 2610.23600 · ▲ 13
- StopLoss: Teaching Agents to Predict Failure and Stop Early University of Waterloo · 2610.25525 · ▲ 12
- Remembering Has a Price: Learning Memory Worth under Resource Constraints Nanjing University · 2610.26453 · ▲ 12
- Competing Learning in Hybrid Architectures Tsinghua University · 2610.26830 · ▲ 12
- UniAVC: Unified Audio-Video Creation via Progressive Training and Asymmetric Scheduling Hong Kong University of Science and Technology, Huawei · 2610.24791 · ▲ 11
- Predicting Privileged Information Effectiveness for On-Policy Self-Distillation University of Michigan, Ann Arbor, Mila, Scale AI · 2610.22512 · ▲ 10
- HandsOn: A Large-Scale Simulation Benchmark for Generalizable Dexterous Manipulation Tsinghua University, Beijing University of Posts and Telecommunications, NatureWill · 2610.26765 · ▲ 10
- A Conditional No-Hair Theorem for Transformers: How Language Models Forget Their Prompt Independent researchers · 2610.24133 · ▲ 9
- World Model Interpretability: From Decoding Representations to Understanding Latent Worlds Shanghai Jiao Tong University, Nanyang Technological University · 2610.25296 · ▲ 9
- TokenGate: Adapting LLMs for Fast and Memory-Efficient Prefill with Learned Token Pruning Stanford University, University of California, Berkeley · 2610.26754 · ▲ 9
- RSIFlow: Adaptive Self-Improvement Workflows with an Evolving Meta-Agent Beihang University, Tsinghua University, Beijing Academy of Artificial Intelligence · 2610.21791 · ▲ 8
- N-Gram Memories at Two Bits: Quantizing Lookup Tables in Large Language Models Dnotitia, KAIST · 2610.21810 · ▲ 8
- HSPR: Recovering Software-Agent Benchmark Scores via Hierarchical Posterior Inference Soochow University, Singapore Management University · 2610.24611 · ▲ 8
- Text-to-Image Generative Modeling Without Language Pretraining Stanford University, Radical Numerics · 2610.26863 · ▲ 8
- Sparse Feature Policy Unlearning Mitigates State Hallucination in Vision-Language-Action Models Chung-Ang University, Georgia Institute of Technology, NAVER AI Lab · 2610.09496 · ▲ 7
- Heterogeneous Agentic Coevolution Meta, University of Virginia · 2610.16078 · ▲ 7
- HERA: Harness-Environment Co-Evolution for Reliable Agentic Abstention University of Washington, University of Leeds, Carnegie Mellon University · 2610.06563 · ▲ 6
- Training Reasoning Models to Generate Diverse Scientific Ideas New York University, Stony Brook University, Salesforce · 2610.22324 · ▲ 6
- QD-Rec: Open-Ended Evolution of Recommendations via Quality-Diversity Sakana AI · 2610.23470 · ▲ 6
- ActionRoPE: action as coordinates for spatially consistent control in planar-camera game world models Tencent, National University of Singapore, Nanyang Technological University · 2610.25861 · ▲ 6
- Any 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale Westlake University, Afari Intelligent Drive, Zhejiang University · 2604.11331 · ▲ 5
- CyberArena: Benchmarking AI Agents in Live Symmetric Attack-and-Defense Combat 2077 AI Open Source Foundation, University of California, Santa Cruz, University of California, Berkeley · 2610.19005 · ▲ 5
- Harness-Delta Attribution: Dissecting Gains From Harness Evolution Amazon, New York University · 2610.22466 · ▲ 5
- SPARC: Spoken Profile Authorship Recognition Corpus University of Southern California, University of Michigan · 2610.25265 · ▲ 5
- Vision-Language-Delta-Action: Learning Compact Deltas as Actionable Transition Interfaces for Robotic Manipulation Harbin Institute of Technology, Shenzhen, Baidu, Shenzhen Loop Area Institute · 2610.25874 · ▲ 5
- Competence-Based Evaluation of Logical Reasoning in Large Language Models University of Illinois Urbana-Champaign · 2610.22894 · ▲ 4
- Token-Set Choice Confounds POPE: A Systematic Audit of Yes/No Extraction in VLM Hallucination Evaluation Sri Krishna College of Technology, Sri Eshwar College of Engineering · 2610.25964 · ▲ 4
- Beyond Maliciousness: Static and Runtime Attack-Surface Profiling of Agent Skills Metano Labs, Purdue University, VulNow · 2610.19264 · ▲ 3
- Measuring Mechanistic Uncertainty in Large Language Models via Causal Robustness University of Colorado Anschutz Medical Campus, University of Wisconsin-Madison, Harvard Medical School · 2610.22636 · ▲ 3
- A Language-Informed Loss for Visual Generation Stanford University · 2610.23653 · ▲ 3
- Legal Guidance in the Wild: How Users Seek Legal Help in Real-World LLM Conversations Maastricht University · 2610.25159 · ▲ 3
- Vision as Proprioception: Estimating Humanoid Robot Pose from Video ETH Zurich · 2610.25288 · ▲ 3
- RealityBench: Simulated Environments to Evaluate a Vision Language Model's Ability to Document Human Physical Work Relantic · 2610.14397 · ▲ 2
- First Passage Attention University of Bergen · 2610.22714 · ▲ 1
Monday, October 19, 2026 · 66
- BizBench: Unveiling Hidden Challenges in Deploying Long-Horizon Agents for Real-World Business Peking University, Ant Group · 2610.20424 · ▲ 492
- Has GPT-6-Astra Achieved Embodied Intelligence? SenseNova, ACE Robotics, Ropedia · 2610.22471 · ▲ 465
- Into the Unknown: Benchmarking AI agents on open scientific research problems Carnegie Mellon University, University of Minnesota, Stanford University · 2610.19159 · ▲ 157
- Learning to Think and Explore in 3D World University of California San Diego, Georgia Institute of Technology · 2610.21783 · ▲ 138 · Deep dive
- Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agents University of California San Diego · 2610.18784 · ▲ 119
- HexaShell: A Thin Harness for Exploring the Limits of VLM Agents as Robot Coding Agents HexaFuture · 2610.23479 · ▲ 118
- LEV: Thinking Fast and Deep with Looped Decision Models Lehigh University, University of British Columbia, Vector Institute · 2610.20951 · ▲ 105 · Deep dive
- MOSS-Image Technical Report MOSI Intelligence, Shanghai Innovation Institute, Fudan University · 2610.23374 · ▲ 94 · Deep dive
- Speculative Sampling in Interactive Video Generation University of Science and Technology of China, Chinese University of Hong Kong, Independent researchers · 2610.20556 · ▲ 84
- Scene-Designer: Learning Design Intelligence for Agentic 3D Scene Generation Zhejiang University, Universal Identification Technology, Hangzhou · 2610.23121 · ▲ 63
- LiveMACE: Process-Aware Evaluation of LLM Agent Capabilities in Evolving Markets National University of Singapore, Fudan University, University of Sydney · 2610.09872 · ▲ 58
- OpenMobile-2: Building Versatile Mobile Agents with Scalable Environments and App-Native Tools Nanjing University, SenseTime, University of Hong Kong · 2610.14157 · ▲ 56
- From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics Research Shanghai AI Laboratory, Shanghai Jiao Tong University, Hong Kong University of Science and Technology, Guangzhou · 2610.21767 · ▲ 54
- Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning Tsinghua University, Tencent · 2610.23181 · ▲ 54
- Xiaomi-Lens-0 Technical Report Xiaomi · 2610.23279 · ▲ 49 · Deep dive
- BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries Fudan University, CocoMatrix, Tsinghua University · 2610.14746 · ▲ 47
- PremiseCheck: Efficient Fact-Checking of Human and AI Peer Reviews Lehigh University, University of British Columbia, Vector Institute · 2610.21046 · ▲ 47
- RAP: A Dataset for Paper-to-Poster Generation via Declarative Markup from Human-Crafted Exemplars Tsinghua University, Meituan Longcat, University of Hong Kong · 2610.23230 · ▲ 36
- Can Visual Generation Expand What Unified Multimodal Models Know? Tsinghua University, Z.ai · 2610.21459 · ▲ 35 · Deep dive
- Lexical Reference Frames: Interpreting LM Residual Streams in Native Semantics Monash University, Chongqing University, Technical University of Munich · 2610.22169 · ▲ 34
- Execution Before Memory: Diagnosing Memory Failures in Video World Models University of Hong Kong, University of Michigan, Ann Arbor, Peking University · 2610.23542 · ▲ 34
- MTEdit: Packed Trajectories Beat Rollouts for Multi-Turn Image Editing SenseTime, Hong Kong University of Science and Technology, Stanford University · 2610.23751 · ▲ 34
- DLite: History-free Speculative Decoding for Agentic AI Shanghai Jiao Tong University, Shanghai Innovation Institute, Tsinghua University · 2610.23787 · ▲ 33
- Lean4AgenticRL: Advancing Agentic RL with Verifiable Multi-level Process Rewards Amazon, University of Illinois Urbana-Champaign · 2610.22448 · ▲ 24
- Consistent Distribution Matching for Data-Free Diffusion Distillation University of British Columbia, Vector Institute · 2610.09221 · ▲ 23
- OmniXRef: A Unified Framework for Multimodal Multi-Reference Video Editing Brown University, Adobe · 2610.22744 · ▲ 22
- OfficeTown: Generative Organizations for Evolving Agent Intelligence Wuhan University, University of California San Diego, South China University of Technology · 2610.23194 · ▲ 21
- Embeddings Are Overpriced? A Frozen 16 MB Vocabulary for Small and Ternary Language Models QLNI · 2610.22428 · ▲ 20 · Deep dive
- DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses PhAI Labs, Chinese University of Hong Kong, University of Michigan · 2610.23839 · ▲ 20
- Belief-Trajectory Energy: Measuring the Path to a Prediction Fudan University, University of Science and Technology of China, Shanghai Innovation Institute · 2610.05114 · ▲ 19
- OPMD: On-Policy Mechanism Distillation Beyond Output Supervision ShanghaiTech University · 2610.14022 · ▲ 18
- Dusting the Phantom: Gaussian Allocation for Fading Dynamic Content Yonsei University, Electronics and Telecommunications Research Institute, POSTECH · 2610.21302 · ▲ 18
- Stuck on Shuffle: On Attribute-Object Binding in Vision-Language Models OriginAI, NVIDIA, Tel Aviv University · 2610.22602 · ▲ 17 · Deep dive
- Towards Looped Models Done Right, Part I: Topology, Input Injection, Recurrent-State Design Institute of Foundation Models, University of Southern California, Carnegie Mellon University · 2610.23036 · ▲ 17 · Deep dive
- MorphAct: Robot Policies Morphing with the Worlds They Act In University of Science and Technology of China, Shanghai Jiao Tong University, National University of Singapore · 2610.20019 · ▲ 16 · Deep dive
- Dream2Discern: Learning from Hard Latent Variants for Runtime Failure Prediction in Multi-Agent Systems Rutgers University, University of Texas at Austin, Amazon · 2610.16071 · ▲ 16
- DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck Google, University of California, Berkeley, Stanford University · 2610.17184 · ▲ 16
- AVC-Reward: Event-Centric Audio-Video Consistency Evaluation Zhejiang University, HiThink Research, Fudan University · 2610.22321 · ▲ 16
- SciΠvot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents University of Maryland, Chinese University of Hong Kong, Gradient Network · 2610.21082 · ▲ 15
- OS-Helix: Unifying GUI Interaction and Code Execution in Native Computer-Using Agents University of Science and Technology of China, Shanghai AI Laboratory, National University of Singapore · 2610.22234 · ▲ 13
- PRISM: PRIvileged Sibling-guided Multi-Agent Topology Design National University of Singapore · 2610.20628 · ▲ 11
- Librarian Agents: Training LLMs to Curate Long-Context Workspaces that Anticipate Future Queries Stanford University, Carnegie Mellon University · 2610.22273 · ▲ 10
- FC-SWE: Failure-Conditioned RL for Long-Horizon Software Engineering Agents NVIDIA, Tsinghua University · 2610.07898 · ▲ 9
- BRACE: Adapting Whole-Body References for Force and Terrain Aware Humanoid Motion Tracking Multiply Labs, Brown University, University of New Mexico · 2610.07052 · ▲ 8
- GlobeBench: Evaluating Text-Based World Model Capabilities for Scalable Environment Simulations Patronus AI · 2610.22708 · ▲ 8
- Failure Recovery via Belief Updates and World Modeling for Long-Horizon Agents University of Maryland, Google, Mohamed bin Zayed University of Artificial Intelligence · 2610.18082 · ▲ 7
- DAGFormer: Global Topology Prediction for Transformer Cross-Layer Connections University of Illinois Urbana-Champaign · 2610.19625 · ▲ 7
- EvoWeb: A Benchmark for Coding Agents on Evolving Web Applications a stealth startup in Tel Aviv, Tel Aviv University · 2610.16402 · ▲ 6
- Recognition Is Not Grounding: Benchmarking Multilingual Financial Evidence Extraction Fin AI, University of Manchester, Georgia Institute of Technology · 2610.19432 · ▲ 6
- Sparse Autoencoders Are a Shared Coordinate System for Symbolic Matching University of Illinois Urbana-Champaign · 2610.19636 · ▲ 6
- Game2Real-Bench: Do Toy Games Reveal and Transfer to Real-World Visual Reasoning? Dalian University of Technology, Nanyang Technological University · 2610.18834 · ▲ 5
- Persistent Planar Memory for Video World Models Tsinghua University, Ant Group, Hong Kong University of Science and Technology · 2610.23795 · ▲ 5
- ReaDy: Recurrent States as Dynamic Context for Speculative Drafting in Hybrid Linear Attention Large Language Models StartLux, University of Chinese Academy of Sciences, University of Hong Kong · 2610.20503 · ▲ 4 · Deep dive
- Robust Explanatory Analysis for Structured Optimization Networks (REASON) in Learning North Carolina State University, Nagoya University · 2610.19378 · ▲ 4
- Skip the Compressor: Coding Agents Fix More Issues When They Pick Whole Files University of Illinois Urbana-Champaign · 2610.21073 · ▲ 4
- Resisting Harmful Instructions through Training-Free Weight Editing University of Illinois Urbana-Champaign · 2610.21189 · ▲ 4
- Selection Is Not Learning: Choosing Model Futures Mohamed bin Zayed University of Artificial Intelligence · 2610.22029 · ▲ 4
- What It Is, Not What It Says: Securing Agentic Browsing Interactions with Web Semantics Washington University in St. Louis · 2610.22697 · ▲ 4
- TokenEconomizer: Experience-Driven Black-Box Cost Control for Coding Agents University of California, Berkeley, Microsoft · 2610.22732 · ▲ 4
- StopWatch: Goal Grounding for Object-goal Navigation Amazon, New York University, University of California, Berkeley · 2610.23841 · ▲ 4
- ARC-SORT: Adaptive, Robust, Compensated Multi-Object Tracking with Interacting Multiple Models Athlitix · 2610.22306 · ▲ 3 · Deep dive
- DisaggKD: Efficient Knowledge Distillation through Teacher-Student Disaggregation and a Fused Loss University of Tokyo, RIKEN Center for Advanced Intelligence Project, SB Intuitions · 2610.17644 · ▲ 3
- DuET: Note-Frame Dual Encoding for Piano Performance-to-Score Transcription National University of Singapore · 2610.21467 · ▲ 3
- AI-Driven Disaster Response and Displacement Monitoring arXiv · 2610.21482 · ▲ 3
- GeoState: Learning View-Decoupled Geometry States for 3D World Modeling Tongji University, Hong Kong Polytechnic University, Tencent · 2610.22343 · ▲ 3
- Relational Learning as World Modeling Technical University of Munich, Munich Center for Machine Learning · 2610.16763 · ▲ 2
Friday, October 16, 2026 · 84
- MJev-Spatial: System One Spatial Intelligence via Native Attention Matching Mind Lab · 2610.22276 · ▲ 545
- OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use Alibaba · 2610.22200 · ▲ 315
- FDE-Gym: The Last Mile of AI Delivery DeepLethe, Tsinghua University, Shanghai Jiao Tong University · 2610.22350 · ▲ 264
- RoTA-Bench: Evaluating 3D Spatial Reasoning through Interactive Projection Alignment Xiamen University, Nanyang Technological University · 2610.17665 · ▲ 196
- Self-Evolving Agent Harness Zhongguancun Academy, Zhongguancun Institute of Artificial Intelligence, University of Science and Technology of China · 2610.18813 · ▲ 195
- Waypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy Distillation Alibaba · 2610.21601 · ▲ 153
- Learning Where to Split: Dynamic Band Splitting for Audio Source Separation Tsinghua University, Chinese Institute for Brain Research · 2610.14993 · ▲ 137
- MAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer Environments Alibaba · 2610.20939 · ▲ 135
- InfiniSplat v2: Analyzing and Improving Feed-Forward 3DGS from Any Views Zhejiang University, Hello TransTech, Shenzhen University · 2610.20341 · ▲ 122
- 3DCodeVerse: Exploring 3D Worlds as Code University of Southern California, Yale University, University of Rochester · 2610.22341 · ▲ 111
- Behavioral Compositionality of Sparse Autoencoder Representations University of Wisconsin-Madison · 2610.21261 · ▲ 104
- Speculative Forcing for Efficient Autoregressive Video Generation KAIST, NAVER AI Lab · 2610.22347 · ▲ 81
- SCOPE: Semantic Bounded Credit Optimization for Prompt Engineering in Text-to-image Generation Xiamen University, University of Rochester, Nanyang Technological University · 2610.22264 · ▲ 76
- VLA-Δ: Learning Physical State Transition for Vision-Language-Action Models Zhejiang University, Central South University · 2610.21306 · ▲ 59
- Each Reward to Its Relevant Tokens: Reward-Specific Token-Level Credit Assignment for Multi-Reward RL University of Minnesota, National University of Singapore, University of Illinois Urbana-Champaign · 2610.21439 · ▲ 59
- Mosaic: Multi-Session Agentic Reinforcement Learning Hong Kong University of Science and Technology, Guangzhou, Huawei, City University of Hong Kong · 2610.22174 · ▲ 54
- CommerceGen-Studio: An End-to-End Framework for Commercial-Grade Main-Image Generation Alibaba · 2610.20603 · ▲ 52
- OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding Hong Kong Polytechnic University, OPPO, Colorado School of Mines · 2610.18829 · ▲ 50
- Intern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image Generation Shanghai AI Laboratory, Shanghai Innovation Institute, University of Sydney · 2610.20818 · ▲ 45
- AutoSFT: An Agentic System for Automated Data Research in Supervised Fine-Tuning Tsinghua University, Beijing University of Posts and Telecommunications, ModelBest · 2610.22194 · ▲ 38
- DeltaStreamer: Towards Human-Like Agentic Streaming Video Understanding Nankai University, ByteDance, IQuest Research · 2610.19849 · ▲ 36
- Orbia: Do Generated Videos Form a Consistent 3D World? National University of Singapore, LightSpeed Studios, University of Hong Kong · 2610.20693 · ▲ 33
- WorldCast: Exploring Intelligent Visual Generation in World State Prediction Tsinghua University, Meituan LongCat, Chinese University of Hong Kong · 2610.20731 · ▲ 33
- Can World and Action Modeling Benefit Each Other under Unified Causal Training? University of Hong Kong, Shanghai Jiao Tong University, Tsinghua University · 2610.21095 · ▲ 31
- Data World: Scaling Data Environments from Scratch Nanjing University, Meituan, Zhongnan University of Economics and Law · 2610.21656 · ▲ 31
- InfoVis: Unified Slide Image Generation and Editing Institute of Automation, Chinese Academy of Sciences, Beijing Academy of Artificial Intelligence, Zhejiang University · 2610.16784 · ▲ 28
- Terminal-RSI: Recursive Self-Improving Terminal Agents via Recursive Self-Rewrite and Policy-Refine Tencent, National University of Singapore, University of Maryland · 2610.22192 · ▲ 28
- EdgeWise: Bringing Long-Horizon Agents to the Edge via Selective Cloud Delegation NVIDIA, Rutgers University · 2610.17560 · ▲ 26
- MLExpert: Towards a Generalist Agentic Model for Autonomous Machine Learning Engineering East China Normal University, Shanghai AI Laboratory · 2610.22344 · ▲ 26
- OmniVerifier-C1: Decomposed Graded Verifier for Rule-Governed Generative World Modeling Tsinghua University, Alibaba, Microcyto · 2610.22355 · ▲ 26
- Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling Shanghai AI Laboratory, Shanghai Jiao Tong University, Fudan University · 2610.22131 · ▲ 25
- Routing Policy Optimization for Online Diffusion Reinforcement Learning National University of Singapore, ByteDance · 2610.22282 · ▲ 24
- Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents Hong Kong University of Science and Technology, Guangzhou, Yinwang, University of Cambridge · 2610.21463 · ▲ 22
- JevAgentHarness: Benchmarking, Deploying and Training Decision Models That Share Control with LLMs University of Illinois Urbana-Champaign, Princeton University, Google · 2610.21412 · ▲ 20
- SimVTA: Simulation-Grounded Video-Trajectory Alignment for Driving World Action Models Shanghai Innovation Institute, University of Hong Kong, Li Auto · 2610.13781 · ▲ 19
- Velox: Spatial Memory Makes World Models Efficient National University of Singapore · 2610.22311 · ▲ 19
- SWE-at-Scale: Scalable Issue Synthesis from Historical Bug Fixes Shanghai Jiao Tong University, North China Electric Power University, Tsinghua University · 2610.21359 · ▲ 18
- Do Better Visual Representations Always Lead to Better End-to-End Autonomous Driving? Southeast University, University of Hong Kong, Shenzhen Loop Area Institute · 2610.09695 · ▲ 17
- ForcingBooster: Adaptive Patch-wise Denoising for Fast Autoregressive Video Diffusion Hong Kong University of Science and Technology, VAST, Macau University of Science and Technology · 2610.12625 · ▲ 16
- Virtual Head Attention Baidu, Peking University · 2610.17959 · ▲ 13
- MFVSR: A Multi-Style and Fine-Grained Benchmark for Video Subtitle Removal National University of Singapore, Tencent, University of Chinese Academy of Sciences · 2610.18461 · ▲ 13
- Athanor: An Agent-Native System for Autonomous Post-Training Research ByteDance, Duke University, University of Chicago · 2610.21055 · ▲ 13
- Learning Self-Orchestration in Decentralized Agent Swarms Harvard University · 2610.22322 · ▲ 11
- FlashMask-3: Efficient and Expressive Mask-Aware Distributed Attention Baidu · 2610.17901 · ▲ 10
- CarryOn: A Dual-Loop Harness for Personalized and Reliable Robot Assistance Hong Kong University of Science and Technology, Chinese University of Hong Kong, SmartMore · 2610.21453 · ▲ 10
- Beyond Policy Rollouts: Embedding-Space Directional Exploration with Anchored Noise for LLM RLVR University of New South Wales, Sydney · 2610.16105 · ▲ 9
- Cowork-Gym: Synthesizing Interactive Work Environments at Scale Alibaba · 2610.17670 · ▲ 9
- SUPRA-TEX: 3D Generation with Super-Aligned Texture Huawei, Shanghai Jiao Tong University, Huazhong University of Science and Technology · 2610.22352 · ▲ 9
- CoDA: Co-Evolution of Data Engines and Action Policies University of Illinois Urbana-Champaign, Chinese University of Hong Kong, Shenzhen, Amazon · 2610.22362 · ▲ 9
- Memory as Weights: Internalizing Long-Term History for Streaming Videos University of Wisconsin-Madison, Peking University · 2610.22070 · ▲ 8
- LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents Simular, University of Wisconsin-Madison · 2610.22280 · ▲ 8
- Calibrated Typed Decisions for Entity, Relation, and Event Extraction University of Manchester, Mayo Clinic, Shanghai University of Finance and Economics · 2610.17113 · ▲ 7
- ANTE: Anticipate, Then Execute -- Learning Internal World Models for Coding Agents from Verified Beliefs University of Maryland, Amazon · 2610.15368 · ▲ 6
- LeanPRM: Lean-Grounded Process Supervision for Natural Language Proof Verification University of Texas at Austin, University of California, Los Angeles, New York University · 2610.17697 · ▲ 6
- SERO: Learning from Executable Oracle Continuations for Long-Horizon Agentic Reinforcement Learning Shanghai AI Laboratory, Shanghai Jiao Tong University, Tongji University · 2610.19397 · ▲ 6
- Mine, Sample, Render: Test-Time Diversity Alignment Of Generative Models Stony Brook University, Microsoft · 2610.21185 · ▲ 6
- Gen4P: Video Generators are 4D Perceivers Nanyang Technological University, University of Oxford · 2610.22371 · ▲ 6
- CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling Salesforce · 2610.06829 · ▲ 5
- Recursive Environment Scaling for General Long-Horizon Agents University of Electronic Science and Technology of China, Hong Kong University of Science and Technology, Chinese University of Hong Kong, Shenzhen · 2610.16397 · ▲ 5
- WorldMemBench: A Comprehensive Benchmark for Memory of Video World Models Zhejiang University, Fudan University, Ant Group · 2610.19063 · ▲ 5
- Copy-Paste Language Models Cornell University · 2610.19247 · ▲ 5
- MMSearch-OPD: Easy Made Hard, Hard Made Learnable for Long-Horizon Multimodal Search Agents Peking University, Hong Kong University of Science and Technology, Guangzhou, Tsinghua University · 2610.22240 · ▲ 5
- Scalable Weak-to-Strong Verifier Hardening for Agent Environments Stanford University, Carnegie Mellon University, Fewshot Corp · 2610.15924 · ▲ 4
- PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference Seoul National University · 2610.19605 · ▲ 4
- Skill-Decoupled On-Policy Expert Distillation for Unified Vision-Language Models Northeastern University, Alibaba · 2610.22217 · ▲ 4
- A team of Gemini 4 Argon agents beats Minecraft Google · 2610.22319 · ▲ 4
- PathFM: Fast Generative Forecasting of Multivariate Trajectories California Institute of Technology, Amazon · 2610.22358 · ▲ 4
- H-DDCM: Hadamard Codebooks for Fast Generative Image and Video Compression Technion, Israel Institute of Technology · 2610.18125 · ▲ 3
- TimeRSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models Georgia Institute of Technology, University of Southern California, Samay AI · 2610.19725 · ▲ 3
- RepForcing: Representation-Space Initialization for Autoregressive Video Distillation Joy Future Academy, Zhejiang University · 2610.20111 · ▲ 3
- One Bias Score, Two Factors: How Mitigations Affect Option Bias in Multiple-Choice Evaluation POSTECH · 2610.20993 · ▲ 3
- Learning to Read Out: Unembedding Dynamics in Language Model Pretraining University of Cambridge, Flower Labs, Isomorphic Labs · 2610.21025 · ▲ 3
- When the KV Cache Is Slower Than the GPU: Scaling Remote KV Tiers for Context-Parallel Prefill Multiverse Computing · 2610.21500 · ▲ 3
- Explaining the Roles of Experts in Mixture-of-Experts LLMs for Personalization University of Copenhagen · 2610.22127 · ▲ 3
- Speech event detector: Timestamped detection of non-verbal vocalizations and background sounds for rich speech transcription arXiv · 2610.14232 · ▲ 2
- PINT: A Unified Model with Shared Predictive Intent for Robot Manipulation Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Zhongguancun Academy · 2610.16355 · ▲ 2
- Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR University of Rochester, Independent researchers · 2610.20461 · ▲ 2
- VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution University of California San Diego, University of Michigan, MIT's Computer Science and Artificial Intelligence Laboratory · 2610.20509 · ▲ 2
- CoWorld Reasoner: Agentic Multimodal Reasoning with the Evolving Video World Model-Driven Harness Hong Kong Polytechnic University, Hangzhou Diagens Biotechnology · 2610.21910 · ▲ 2
- EVA-Cap: Optimizing Audiovisual Video Captioning via Event-Centric Alignment Fudan University, OPPO · 2610.22186 · ▲ 2
- SimForcing: Distilling Simulation Motion Priors into Real-Domain Robot World Models Peking University · 2610.06598 · ▲ 1
- AI-Driven, Timing-Aware Power Optimization arXiv · 2610.18890 · ▲ 1
- Diminishing Returns of Data in Sparse Autoencoders: Functional Behavior does not Guarantee Feature Reproducibility IBM, Indian Institute of Science, University of Virginia · 2610.20680 · ▲ 1
- PAIR: Bridging Perception and Action in Vision-Language-Action Models University of Maryland · 2610.09016
Thursday, October 15, 2026 · 88
- Any Time, Any Space: A World-State-Centric Framework for Agentic Video Generation Zhejiang University, Monash University, University of California, Berkeley · 2610.20571 · ▲ 344
- Player-1: Learning to Play from Synthetic Experience National University of Singapore, Tencent, Nanyang Technological University · 2610.20584 · ▲ 336
- Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation Nanjing University, JD.com · 2610.13521 · ▲ 243
- Doc2Harness: Benchmarking LLMs on Long-Horizon Harness Engineering of Coding Agents Institute of Computing Technology, Chinese Academy of Sciences, Tsinghua University, Peking University · 2610.20344 · ▲ 106
- More Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement Learning Zhejiang University, Ant Group · 2610.19789 · ▲ 83
- Recovering Reliable Supervision via Critique-Guided On-Policy Distillation Brigham Young University, University of North Carolina at Chapel Hill · 2610.20572 · ▲ 80
- Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders Nanyang Technological University · 2610.17771 · ▲ 70
- SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm Huawei, Leiden University, University of Groningen · 2610.20431 · ▲ 69
- LatentLoop: Scaling Looped MoE Models with Asymmetric Compute Allocation Shanghai Jiao Tong University, Shanghai AI Laboratory, University of Science and Technology of China · 2610.19975 · ▲ 67
- Cortex: Consolidating Context into Parametric Memory Zhejiang University, ByteDance, Tsinghua University · 2610.20399 · ▲ 67
- ActionOPD: Distilling Decisions, Not Tokens Tsinghua University · 2610.19820 · ▲ 59
- Seeing, Maintaining, and Learning: The Evolution of Multimodal Agent Memory Rutgers University, University of Notre Dame, University of Michigan · 2610.19366 · ▲ 51
- LensErase: A Foundation Model for Erase Anything Shanghai Jiao Tong University, Nankai University, University of Science and Technology of China · 2610.18072 · ▲ 45
- KV4-RL: Unlocking Faster Asynchronous Rollouts for RLVR with INT4 KV Cache Quantization KAIST, NAVER Cloud, AITRICS · 2610.14448 · ▲ 36
- Towards Agentic Robots: Model, Data, Environment, and Harness Chinese University of Hong Kong, Fudan University, NovaXBot · 2610.20442 · ▲ 34
- CLASHarness: Benchmarking Multi-Agent Conflict in Computer Use via a Controllable and Verifiable Generation Pipeline Google, Emory University, University of Southern California · 2610.19128 · ▲ 33
- Hybrid Reasoning Agents Stanford University, Together AI, University of Illinois Urbana-Champaign · 2610.18861 · ▲ 29
- HumanoidToolBench: Benchmarking Humanoid Tool Use from Selection to Mobile Execution Seoul National University, University of Massachusetts Amherst, Google · 2610.02089 · ▲ 28
- RuleTrek: Benchmarking Agentic Rule Reasoning in Complex Scenarios Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Beijing Academy of Artificial Intelligence · 2610.19698 · ▲ 28
- The Thinking Leakage in Reinforcement Learning for LLMs King Abdullah University of Science and Technology, Shanghai Jiao Tong University · 2610.14270 · ▲ 27
- APT: Feedback-Driven Agentic Post-Training for Text-to-Image Generation University of Science and Technology of China, Microsoft, University of Sydney · 2610.18097 · ▲ 22
- Super-Resolution in The Right Latent Space: A Frozen Vision-Foundation Substrate University of California, Berkeley, Alibaba, University of Michigan, Ann Arbor · 2610.04781 · ▲ 21
- APEX-Harm: Evaluating Agent Safety in Shared Professional Workspaces King Abdullah University of Science and Technology, University of Oxford, CAMEL-AI · 2610.13623 · ▲ 21
- τ-Emotion-Bench: Exposing the Interaction Fragility of LLM Agents through Real-World User Emotions Arizona State University · 2610.19560 · ▲ 21
- DecisionRobust: Benchmarking Robustness for JEV Style Decision Models Across Text and Vision Pennsylvania State University · 2610.19118 · ▲ 20
- Apertus 1.5: From Open Text Models to Open Multimodal Reasoners ETH Zurich, EPFL, University of Applied Sciences Western Switzerland, Valais-Wallis · 2610.15542 · ▲ 19
- Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models Nanyang Technological University, Yale University, University of Manchester · 2610.19696 · ▲ 19
- A Diagnostic Study of World-Space Human Recovery Under High Camera Dynamics National Yang Ming Chiao Tung University, Boston University · 2610.19729 · ▲ 17
- LabInstruct: Benchmarking Situated Instructional Video Generation for Lab Procedures Central South University, National University of Singapore, Beijing University of Posts and Telecommunications · 2610.14020 · ▲ 16
- Long-Horizon RL Needs Muon Renmin University of China, Duke University, Institute of Automation, Chinese Academy of Sciences · 2610.20256 · ▲ 16
- ProgAgent: Learning Tool Orchestration through Programmatic Training Hong Kong University of Science and Technology, Beijing University of Posts and Telecommunications, Nanyang Technological University · 2610.14774 · ▲ 15
- SkillMarker: A KGW-Based Watermarking Method for Agent Skill Ownership Verification Huawei, Harbin Institute of Technology, Shenzhen · 2610.15725 · ▲ 15
- RAR: Rate-Adaptive Rollout for Time-Consistent Interactive World Models Hong Kong University of Science and Technology, Hong Kong University of Science and Technology, Guangzhou, Lightspeed · 2610.16302 · ▲ 15
- Are We Testing Too Much? A Survey of Benchmark Compression for Efficient Evaluation of Large Language Models Sun Yat-sen University · 2610.18248 · ▲ 15
- StartLux-Decision: AutoThink Decision Models with Reinforcement Learning for Calibrated Deliberative Decisions StartLux, University of Hong Kong, Southern University of Science and Technology · 2610.17309 · ▲ 14
- Efficient Scene-Text Refinement via Pixel-Space Grouped Image Generation Zhejiang University, Alibaba, Peking University · 2610.18284 · ▲ 14
- Joint Subspace Sparse Attention for Efficient Long-Context Decoding Seoul National University · 2610.19026 · ▲ 14
- GroundGUI: GUI Grounding by High-Quality Data and Distillation-to-Reinforcement Learning Prentis AI, Westlake University, Titan Holdings · 2610.19100 · ▲ 14
- FreeSolo: Linear-time Sparse Attention with Zero-Shot Length Extrapolation National University of Singapore, University of Toronto, Sea AI Lab · 2610.17191 · ▲ 13
- Diagnosing and Correcting Reward Hacking in Verifiable Instruction Following Yonsei University, Furiosa AI, University of Washington · 2610.19525 · ▲ 13
- SafeDriveVLA: Navigation-Conditioned World Model Dreaming for Conflict-Aware End-to-End Autonomous Driving University of California, Irvine, University of California San Diego, University of Cincinnati · 2610.20609 · ▲ 12
- LLM-to-Flow Distillation: Teaching Natural Language Goals to Scientific Generators ETH Zurich, ETH AI Center, Vienna University of Technology · 2610.17246 · ▲ 11
- Rethinking Projector Training in Multimodal LLMs EPFL, University of Basel, Lucerne University of Applied Sciences and Arts · 2610.18598 · ▲ 11
- Self-Conditioning Enables Deep Sequential Reasoning in Flow Matching Shanghai Qi Zhi Institute, Tsinghua University, Xiongan AI Institute · 2610.19628 · ▲ 11
- What Makes a Social Agent's Interactions Effective? Learning from Verifiable Outcomes in Paired Worlds Alibaba, University of Tokyo · 2610.20195 · ▲ 11
- History Replays Itself: Unleashing Long-Context Capability in Linear Attention King Abdullah University of Science and Technology · 2610.15715 · ▲ 10
- ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty Florida State University, University of Tokyo, RIKEN Center for Advanced Intelligence Project · 2610.17016 · ▲ 10
- BigCell: Generating Gigapixel Whole-Slide Images Stony Brook University, Argonne National Laboratory, University of Chicago · 2610.14261 · ▲ 9
- Acting Through Registers Shanghai Jiao Tong University, Institute of Automation, Chinese Academy of Sciences, Humanoid Robot Shanghai · 2610.17638 · ▲ 9
- From Failures to Learnable Worlds: Policy-Conditioned Environment Generation for Navigation Agents UNC Chapel Hill · 2610.19420 · ▲ 9
- TextVerifier: Seeing Beyond Recognition to Verify Chinese Text Rendering Peking University, Alibaba, Zhejiang University · 2610.19551 · ▲ 9
- Glance Forcing: Making Bidirectional Diffusion Autoregressive at a Glance Wuhan University, University of Electronic Science and Technology of China, Sun Yat-sen University · 2610.20220 · ▲ 9
- The COUNT: Counting N-grams Unlocks Scaling on Host Memory and Training-Free Adaptation University of Washington, Allen Institute for AI · 2610.20456 · ▲ 9
- Continuum Attention University of Luxembourg · 2610.20586 · ▲ 9
- VITAL: From Visual Motion to Precise Action Horizon Robotics, Shanghai Innovation Institute, Fudan University · 2610.20604 · ▲ 9
- SWEeper-Bench: Can Agents Discover Bugs in Interactive Software? Princeton University · 2610.14096 · ▲ 8
- Exploring the Limits of Combinatorial Generalization in Embodied Manipulation Shanghai Jiao Tong University · 2610.15425 · ▲ 8
- Rethinking LLM-Guided Visual Question Answering: Does Deeper Knowledge Integration Really Help? Independent researchers · 2610.17079 · ▲ 8
- Diff Diffusion: Predicting Edit Paths with Diffusion Models for Efficient Code Editing ByteDance · 2610.17831 · ▲ 8
- Training LLMs for Multi-Agent Collaboration with Verifiable Information Rewards Arizona State University, University of Houston, Ohio State University · 2610.17893 · ▲ 8
- PreciseCoder: Faithful and Generalizable Debugging Beyond Unit-Test Rewards University of Southern California, University of Chicago, Google · 2610.18755 · ▲ 8
- Generative World Simulation with DAO and JING XGEN Team · 2610.19726 · ▲ 8
- VASPO: Self-Distillation with Privileged Value Estimation University of Science and Technology of China, iFLYTEK Research · 2610.20516 · ▲ 8
- Latent Compaction with Test-Time Training Tsinghua University · 2610.20582 · ▲ 8
- SyncReward: Human-Aligned Synchronization Reward for Audio-Visual Generation Renmin University of China, Tencent · 2610.16338 · ▲ 7
- Scaling Online Optimal Transport Minibatches for Flow Matching University of California, Berkeley · 2610.18838 · ▲ 7
- MA-Bench: Measuring Operational Planning in Language Models Scale AI, Data Machines · 2610.19138 · ▲ 7
- MimicX: Policy-in-the-Loop Supervision Refinement for Video-Driven Humanoid Motion Tracking Hong Kong University of Science and Technology, Guangzhou · 2610.09055 · ▲ 6
- NavGPT-3: Harnessing Context in a Hierarchical Navigation Runtime Adelaide University, Metacognition, Australian National University · 2610.10787 · ▲ 6
- Jag: Direct Box Prediction for Efficient Visual Grounding University of Chinese Academy of Sciences, State Key Laboratory of Communication Content Cognition, Peng Cheng Laboratory · 2610.13239 · ▲ 6
- Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation Stanford University, Together AI · 2610.14162 · ▲ 6
- When Left Is Right? Camera-Motion Understanding beyond Screen Hong Kong University of Science and Technology, Guangzhou, Tencent, École Polytechnique · 2610.17242 · ▲ 6
- Choosing LLMs for Strategic Agent Interactions University of Exeter, IBM · 2610.18354 · ▲ 6
- To Align or Not To Align: Check Your COMPASS Before You Train EPFL, University of Basel, Lucerne University of Applied Sciences and Arts · 2610.18638 · ▲ 6
- AgentSafeEvo: Adversarial Co-Evolution for LLM Agent Safety Shanghai AI Laboratory, Fudan University, Shanghai Jiao Tong University · 2610.19422 · ▲ 6
- Physical Time as a Missing Dimension of Robot Policy Memory Dynamics New York University Shanghai, University of Electronic Science and Technology of China · 2610.20524 · ▲ 6
- G^3VLA: Geometric Inductive Bias for Vision-Language-Action Models New York University Shanghai, Technical University of Denmark, Mohamed bin Zayed University of Artificial Intelligence · 2606.24472 · ▲ 5
- Vela: Scaling Vision-Language-Action Models with Adaptive Action Curve Parametrization Fudan University, Shanghai Innovation Institute, Chinese University of Hong Kong · 2610.05230 · ▲ 5
- Unlocking Structured Pruning for Reasoning Language Models via Attention-Preserving MLP Trimming Seoul National University, New York University, University of Southern California · 2610.19121 · ▲ 5
- StreamErase: Towards Ultra-Fast Streaming Video Erasing Wuhan University, Alibaba, Hong Kong University of Science and Technology · 2610.20137 · ▲ 5
- Your Transformer Is Secretly Elastic University of Amsterdam, TNO · 2610.20588 · ▲ 5
- CoTranslator: Reverse to Forward Chain-of-Thought Translator for Reasoning Distillation Tsinghua University, Ant Group, Xiaomi · 2610.20828 · ▲ 4
- Cortex: Measurement-Gated Evolution and Pruning of Coding-Agent Skills and Rules Independent researchers · 2610.14216 · ▲ 3
- VLX-Go: Robust Embodied Tracking through Multi-Stage Learning and On-Policy Refinement Om AI Research, Binjiang Institute of Zhejiang University · 2610.19361 · ▲ 3
- Lluna: Collaborative Unimodal Models for Missing-Modality Inference College of William and Mary, Carnegie Mellon University, Microsoft · 2610.20526 · ▲ 3
- The Agent Knew: Internal States Forecast Unauthorized Actions and Stop Them, and the Lies That Follow University of Southern California · 2610.17301 · ▲ 2
- Looking Complete Is Not Being Grounded: Evaluating LLM Sitreps Against Human-Authored Situation Reports LexisNexis · 2610.18762 · ▲ 2
- Drive3R: Block-wise Streaming Geometric Foundation Models for Autonomous Driving Shanghai Jiao Tong University, Didi Chuxing, Chinese University of Hong Kong, Shenzhen · 2610.15112 · ▲ 1
Monday, October 12, 2026 · 60
- S-Space: Exploring Spatial Workspace in Multimodal Language Models MirroS · 2610.13527 · ▲ 294
- DiffusionART: Efficient, Stable and Scalable Reinforcement of Diffusion Models National University of Singapore, Kuaishou, Hong Kong University of Science and Technology · 2610.12572 · ▲ 148
- SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments Shanghai Jiao Tong University, Zhejiang University · 2610.12611 · ▲ 107
- Reflex: Visually Grounded Reactive Humanoid Control University of Washington, Stanford University, University of Southern California · 2610.12772 · ▲ 88
- EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents Peking University, Beijing Institute for General Artificial Intelligence, University of Hong Kong · 2610.13021 · ▲ 73
- VisionWeave: Weaving Elastic Visual Representations as a Native Capability of MLLMs University of Science and Technology of China, Alibaba, Chinese University of Hong Kong · 2610.07987 · ▲ 68
- Hear the World in Stereo: Learning Dynamic Spatial Correspondence for Immersive Joint Video-Audio Generation Xidian University, vivo BlueImage Lab at vivo Mobile Communication · 2609.38748 · ▲ 65
- ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents Xiamen University, Washington University in St. Louis, Institute of Automation, Chinese Academy of Sciences · 2610.13395 · ▲ 54
- Work2Skill: Capturing Human GUI Experience for Reusable Agent Skill Induction Shanghai Jiao Tong University, Fudan University, Nanyang Technological University · 2610.12984 · ▲ 51
- RSI-Anything: Unlocking Recursive Self-Improvement with World Models PhAI Labs, Chinese University of Hong Kong, University of Macau · 2610.13912 · ▲ 46
- SepGen: Multi-Stem Audio-Video Separation and Generation in a Single Model Tel Aviv University, Google · 2610.11361 · ▲ 45
- Continually Post-training Small Models via Synthetic Self-distillation KAIST, LG AI Research · 2610.13113 · ▲ 45
- SoloQ: Calibration-Free Quantization for Diffusion Language Models University of Southern California, Yale University, Korea University · 2610.07121 · ▲ 39
- Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition Tsinghua University · 2610.12341 · ▲ 38
- PolyOCR-Venus: Unified OCR Foundation Models for Text-Centric Visual Intelligence Ant Group · 2609.37712 · ▲ 36
- Learning Representations to Guide Pixel Diffusion Transformers Sun Yat-sen University, Alibaba, University of Hong Kong · 2610.12978 · ▲ 31
- Stateless Language Agents: Scaling Long-Horizon Automated Research Stanford University, Carnegie Mellon University, University of Washington · 2610.07625 · ▲ 29
- Contextual Correction of Parallel Drafts for Speculative Decoding a stealth startup in Tel Aviv, Tel Aviv University, Ben-Gurion University · 2610.12570 · ▲ 29
- InstantFusion: A Shared Latent Interface for Heterogeneous Image Generators Fudan University, Shanghai AI Laboratory, Shanghai Jiao Tong University · 2610.13944 · ▲ 25
- Judging in Latent Space: Efficient Generative Reward Modeling via Semantics-Preserving Compression Soochow University, University of Cambridge, Chalmers University of Technology · 2610.09788 · ▲ 23
- When Do We Need On-Policy Distillation? Distilling on Offline Student Rollouts Is Often Better NVIDIA, University of California, Los Angeles · 2610.11291 · ▲ 23
- Recurrent Harness for Agentic Continual Learning Massachusetts Institute of Technology, Meta, Yale University · 2610.12668 · ▲ 23
- Towards Physically Grounded Articulated 3D Assets: A Multi-Agent Pipeline and Benchmark Nanyang Technological University, ETH Zurich, University of Oxford · 2610.13947 · ▲ 23
- Inducing Process Supervision from Outcome-Only Reinforcement Learning Renmin University of China, Tsinghua University, University of Electronic Science and Technology of China · 2609.36641 · ▲ 21
- MetaKernelBench: Measuring GPU Kernel Knowledge Transfer Beyond Code Westlake University, Shanghai Jiao Tong University, Zhejiang University · 2610.05014 · ▲ 21
- SWE-Duel: Towards a Self-Scaling Arena for Coding Agents Singapore Management University · 2610.12569 · ▲ 21
- GraphDecisionBench: Benchmarking System One Models on Graph Tasks Microsoft, Beijing University of Technology · 2610.06354 · ▲ 19
- TouchScale: 500 Hours of Human Vision and Touch for Visual-Tactile Learning Texas A and M University, Google, Carnegie Mellon University · 2610.10288 · ▲ 18
- From Given to Gathered Evidence: Agentic Learning for Longitudinal Medical Reasoning Durham University, Alibaba, Hupan Laboratory · 2609.39566 · ▲ 17
- DataVista: Diagnosing Multimodal LLMs on Data Video Understanding Hong Kong University of Science and Technology, Guangzhou, Zhejiang University · 2610.11993 · ▲ 17
- Surfing the Activation Wave: Analyzing the Information and Geometry of LLM Hidden States for Speculative Decoding Qualcomm AI Research, Stanford University · 2610.12939 · ▲ 17
- Universal Textual Teaching for LLMs Westlake University · 2610.12114 · ▲ 16
- The value of information and the emergence of communication in evolved e-puck robots Universidad Veracruzana · 2610.13118 · ▲ 16
- Rethinking Relative Policy Differences Transfer of On-Policy Distillation Zhejiang University, Shanghai AI Laboratory, Fudan University · 2610.13205 · ▲ 16
- Video-RSI: Recursive Self-Improvement of Video Understanding Agents via Harness Evolution Tsinghua University, Harbin Engineering University · 2609.37950 · ▲ 14
- Humanity's Sixth Sense: Benchmarking Intuitive Visual Reasoning in Multimodal Models Scale AI, Elorian, University of California, Santa Cruz · 2610.08966 · ▲ 14
- Coding-Agent Benchmarks Should Match Their Users' Task Flows JetBrains Research · 2610.09633 · ▲ 14
- PreSTO: Predictive Subtree Prefetching for Fast LLM Power Sampling University of Texas at El Paso, Georgia Institute of Technology, Morgan Stanley · 2610.13171 · ▲ 14
- When Do Biological Reasoning Models Use Their Biological Inputs? Harvard University, Massachusetts Institute of Technology · 2610.00898 · ▲ 13
- MixTok: Adaptive Visual Tokenization with Global 1D Tokens and Sparse 2D Refinement Zhejiang University, Hong Kong University of Science and Technology, University of Adelaide · 2610.13181 · ▲ 13
- AMLworld-Compact: Importance-Weighted Downsampling for Cost-Effective LLM Evaluation and Error Diagnosis SCB DataX at SCBX Group, Independent researchers · 2610.13627 · ▲ 13
- One Model to Loop Them All: Elastic Latent Recurrence for Language Models NVIDIA · 2610.12582 · ▲ 12
- Less from More: Reinforcing Sparse Video Reasoning from Dense References University of Amsterdam · 2610.10893 · ▲ 10
- The Life Cycle of a Massive Activation: Stochastic Birth, Weight-Decay-Driven Growth, and Competitive Consolidation Aimpoint Digital Labs, Mistral AI, New York University · 2610.00423 · ▲ 8
- Distributionally Robust Mixture-of-Experts Training New York University, NYU Shanghai · 2610.07207 · ▲ 8
- QuantMLA: Function-Aligned Dual-Path Quantization for Low-Bit MLA KV Caching University of Hong Kong, Meituan LongCat Team, South China University of Technology · 2609.36760 · ▲ 7
- ThinkDrag: Semantic Drag-Based Image Editing with Visual Reasoning Nanyang Technological University · 2610.13316 · ▲ 7
- BudgetPix: Compute-Adaptive Tokenization for Pixel-Space Image Diffusion University of Illinois Urbana-Champaign, Google · 2610.12307 · ▲ 6
- Mental-Models for Multi-Agent Systems University of California San Diego · 2610.12453 · ▲ 6
- Encore: Few-Shot Agentic Discovery of Manipulation Strategies Massachusetts Institute of Technology, twenty seventy-seven AI Open Source Foundation, Texas A and M University · 2609.37359 · ▲ 3
- Which Language Should a Skeleton Speak? Language Choices in Multilingual Reasoning Electronics and Telecommunications Research Institute, KAIST, Dankook University · 2610.09607 · ▲ 3
- Clinician use of language models diverges from how the models are evaluated NYU Langone Health, New York University, Washington University School of Medicine · 2610.11069 · ▲ 3
- Self-Preservation in Language Models New York University, Google, Cornell University · 2610.12602 · ▲ 3
- SAGE: Salient Factor Discovery and Generation with Visual Foundation Representations University of Hong Kong, Boston College, University of Virginia · 2609.39635 · ▲ 2
- When to Rethink: Learning Multi-Perspective Self-Verification for Vision-Language Models University of Leicester, Nanjing University of Posts and Telecommunications, University of North Carolina at Chapel Hill · 2610.07018 · ▲ 2
- Few-Step Generation via Data-Space Iteration ByteDance · 2610.12102 · ▲ 2
- Video-Conditioned Generative Joint 2D-3D Hand Motion Recovery University of Hong Kong, University of California, Berkeley · 2610.10512 · ▲ 1
- GATOR: Generative and Agentic 3D Object Reconstruction From Casual Images NVIDIA, Simon Fraser University · 2610.11215 · ▲ 1
- MAMHOI: Factorizing Scene-Aware Human-Object Interaction through Affordances Nanyang Technological University · 2610.12416 · ▲ 1
- Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration Seoul National University, University of Maryland, All Purpose AI · 2610.12470 · ▲ 1
Friday, October 9, 2026 · 72
- AgentGarten: Code Worlds for Evolving Agents MirroS · 2610.12374 · ▲ 260
- Learn2Play Bench: How Well Do LLM Agents Learn from Experience in Unfamiliar Environments? National University of Singapore · 2610.08215 · ▲ 141
- TokenRouter: Efficient Serving System for Token-Level LLM Routing Tsinghua University, Carnegie Mellon University · 2610.12242 · ▲ 130
- From Traces to Agentic Worlds: Agentic Language World Models for Interactive Environment Simulation Nanyang Technological University, Hong Kong Polytechnic University · 2610.06100 · ▲ 107
- MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement Xiaomi · 2610.11959 · ▲ 77
- SuperNav: An Agentic Navigation System for Any Task in Any Scene Zhejiang University, Shenzhen University, Causa Robotics · 2610.12126 · ▲ 75
- Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction KAIST · 2610.12299 · ▲ 57
- In-context Robot Learning Made Simple: A Democratized Recipe for Manipulation Tasks arXiv · 2609.38173 · ▲ 51
- REMORY: Learning Residual Memory for Context Compaction Shanghai Academy of AI for Science, Fudan University, Shanghai Jiao Tong University · 2610.11287 · ▲ 50
- U-Space: Uncovering When and Why Uncertainty Arises in Language Models Technical University of Darmstadt, University College London, University of Lübeck · 2610.09087 · ▲ 46
- OuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3D Cinemagraphs National Yang Ming Chiao Tung University, Alaya Lab · 2610.12461 · ▲ 44
- MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers Fudan University, Tencent, Shanghai Innovation Institute · 2610.06801 · ▲ 39
- Beyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence Matching University of Hong Kong, ByteDance, Zhejiang University · 2610.12421 · ▲ 39
- DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training Institute of Automation, Chinese Academy of Sciences, Alibaba · 2610.12468 · ▲ 39
- Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks University College London, Huawei, University of Liverpool · 2610.11794 · ▲ 37
- TestPrism: Rethinking Test Evaluation Beyond a Single Reference Nanjing University, Hong Kong Polytechnic University, Tencent · 2610.12289 · ▲ 37
- SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference Westlake University, Ant Group · 2610.12327 · ▲ 32
- Post-Training Frontier Text-to-Image Models by Composing Preference and Rubric Rewards Arena Intelligence, University of California, Los Angeles · 2610.02967 · ▲ 31
- Foundations of Large Language Models arXiv · 2501.09223 · ▲ 29
- OneSearch-VL: Unified Multimodal Deep Research Agent for Image and Video Beihang University, Chinese University of Hong Kong, NTU · 2610.12419 · ▲ 25
- LEGO: A Lifting-Free Approach for Exocentric-to-Egocentric Video Generation GenGenAI · 2610.12442 · ▲ 25
- Reasoning-Informed Visual Editing Shanghai Jiao Tong University, Southeast University, South China University of Technology · 2610.12343 · ▲ 22
- Embodied Turing Machines: Stateful Code for Robot Recursive Self-Improvement Nanyang Technological University, Ropedia · 2610.12369 · ▲ 22
- What Did the Agent Actually Do? Evidence-Grounded Oversight for Long-Horizon Agents Renmin University of China, Kuaishou · 2610.06406 · ▲ 21
- VibeEdit: Image Editing with Canvas Instructions University of Sydney, Microsoft, Fudan University · 2610.12229 · ▲ 19
- SparseEngine: Sparse-First Inference Engine arXiv · 2609.39068 · ▲ 18
- USDCraft: Geometrically Grounded Programmatic Modeling of Articulated 3D Assets for Simulation NVIDIA, NTU, Peking University · 2610.11322 · ▲ 18
- Pumpire: Unified Benchmark for Metric Distance Estimation Zhejiang University, Zhejiang University of Technology, Shanghai Jiao Tong University · 2610.12423 · ▲ 18
- SanSi: A Looped Typed Decision Model for System 1.5 Thinking University of Minnesota · 2610.07730 · ▲ 17
- Opera: A Verbal Critic Framework for Long-horizon Coding Agents Salesforce, Rutgers University, Lehigh University · 2609.33987 · ▲ 16
- A GPU-Parallel Framework for Heterogeneous Multi-Task Reinforcement Learning NVIDIA, Hong Kong University of Science and Technology, Guangzhou, Tsinghua University · 2606.03335 · ▲ 15
- ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills Zhejiang University · 2610.12403 · ▲ 15
- OmniCapBench: A Deep-Structured Evaluation Framework for Fine-Grained Audio-Visual Captioning Tencent, Nanyang Technological University, Northumbria University · 2610.12458 · ▲ 15
- Do LLMs Understand Sequential Structure? A Controlled Study of Inference and Generation University of Illinois Urbana-Champaign, Stony Brook University, National Chengchi University · 2610.04977 · ▲ 14
- ReSPO: Reshaped Sequence Policy Optimization for Gradient Starvation in Off-Policy Learning University of California, Los Angeles · 2609.35433 · ▲ 13
- SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces Alibaba · 2610.11366 · ▲ 13
- WorldGuide: Goal-Directed Video World Model for Procedural Task Execution Mohamed bin Zayed University of Artificial Intelligence · 2610.12459 · ▲ 13
- Retrieval-Centric Deep Learning in Growing Nonparametric Neural Networks Google, Massachusetts Institute of Technology, Yale University · 2610.03858 · ▲ 12
- V-CoLA: Vision Token Compression with Linear Attention Alibaba · 2610.11251 · ▲ 12
- From Prompting to Composing: A Spatial Canvas Interface for Poster Generation Fudan University, Microsoft, University of Sydney · 2610.12230 · ▲ 12
- SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models Zhejiang University · 2610.12402 · ▲ 11
- A Closer Look at Agentic BBO: Benchmarking LLM Agents for Black-Box Optimization Nanjing University · 2610.12183 · ▲ 10
- Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict Johns Hopkins University, New York University · 2610.12360 · ▲ 10
- BrickBench: Evaluating Agentic Brick Design Stanford University, Max Planck Institute for Intelligent Systems, Inria · 2610.12452 · ▲ 10
- Can AI Agents Make Open-Ended Scientific Discovery? Evidence from Station DualverseAI, University of Hong Kong, University of Cambridge · 2610.08927 · ▲ 9
- Chaos in the Text: Revealing the Modality Preference in Mixed-Modality Retrievers University of Chinese Academy of Sciences, Northeastern University, Tsinghua University · 2610.11816 · ▲ 9
- Mara Chain: Rethinking Failure as a Stepping Stone for AI System Auto-Evolution Ant Group, Beijing Intelligent Game and Decision Lab, Beijing Defense Innovation Institute · 2609.35855 · ▲ 8
- Scaling to Tens of Thousands of Test-Time Iterations with Loop-Native Attention Residuals Hong Kong Polytechnic University, Independent researchers, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences · 2610.11570 · ▲ 8
- Distilling Routed 3D Privilege for Spatial Reasoning in Vision-Language Models Zhejiang University · 2610.12355 · ▲ 8
- SpaceFlow: Locally Controllable 3D Generation ETH Zurich, Stanford University, Microsoft · 2610.12399 · ▲ 8
- One Block, Multiple Depths: Recurrent Vision Transformers with Depth-Programmed Experts Samsung, Technical University of Iasi, Queen Mary University of London · 2610.12448 · ▲ 8
- You Changed Your Mind, The Model Didn't: Demystifying Intent in Multi-Turn Dialogue Hong Kong University of Science and Technology, Tencent · 2610.06496 · ▲ 7
- CARE: Certifying Acceleration for Vision-Language-Action Inference University of Maryland, University of Oxford, Google · 2610.08917 · ▲ 7
- On-Policy Distillation Teaches New Skills but Not New Knowledge Hong Kong University of Science and Technology · 2610.09639 · ▲ 7
- MIRA: A Musical Intent Refinement Agent for Aligning Text-to-Music Generation with User Intent Shandong University, University of Science and Technology of China, Central Conservatory of Music · 2610.10355 · ▲ 7
- Synthesis Through Simulation: Generating Coherent Enterprise Data via Scalable Agent-System Interaction SAP Labs · 2610.10549 · ▲ 7
- Incremental Open-Ended Deep Research with Structured Harness Xiaohongshu, Zhejiang University · 2610.11566 · ▲ 7
- SpecFold: Folding Multi-Branch Redundancy for Faster Speculative Decoding in Diffusion Language Models Georgia Institute of Technology · 2610.04875 · ▲ 6
- EDiS: Edge Disjoint Subgraph Sparsification Framework for Graph Neural Networks University of Texas at Dallas, Pacific Northwest National Laboratory, Adobe · 2610.09059 · ▲ 6
- Frozen Models, Evolving Expertise: Model-Agnostic Learning from Deployment Experience for Multimodal Medical AI University of Maryland, NVIDIA · 2610.09146 · ▲ 6
- MARGIN: Runtime Confidence Calibration for Multi-Agent Foundation Model Coordination Ericsson · 2605.22949 · ▲ 5
- Learning to Steer, Steering to See: Unveiling the Geometry of RLVR in Large Language Models via Trainable Vectors University of Science and Technology of China, Tencent, Beijing University of Posts and Telecommunications · 2609.34344 · ▲ 5
- Investigating the Role of Reasoning-Language Alignment in Monolingual Retrieval-Augmented Generation Johannes Gutenberg University Mainz, University of Colorado Boulder · 2610.03136 · ▲ 5
- Incidental information contaminates patient notes and disrupts clinical reasoning in large language models arXiv · 2610.08585 · ▲ 5
- SPW-Nav: A Streaming Panoramic World Model for Language-Guided Navigation Alaya Lab, Peking University, Institute of Science Tokyo · 2610.08941 · ▲ 5
- Skill Constellations: Tracing the Supply Chain of Agent Skills on GitHub University of British Columbia, Okanagan · 2610.11169 · ▲ 5
- The Lattice of Transition Laws University of Pennsylvania · 2610.11216 · ▲ 5
- Predicting Cable Dynamics with Physical Attention Bias Tel Aviv University, Meta · 2610.11975 · ▲ 5
- SatNav: A Scalable Benchmark for Long-Horizon UAV Vision-Language Navigation from Satellite Imagery Hong Kong University of Science and Technology, Guangzhou, International Digital Economy Academy, Hong Kong University of Science and Technology · 2609.31507 · ▲ 4
- Behavioral Persistence and Incomplete Functional Transfer of Co-evolved Communication in Evolutionary Robotics Universidad Veracruzana · 2609.38527 · ▲ 4
- TerraVis: Towards Evaluation of World-Grounded Visual Consistency in Text-to-Image Generation via MLLM Workflows Adelaide University, xAI, Responsible AI Research Centre · 2610.02959 · ▲ 4
- Evaluating the Transfer of Co-Evolved Communication from 2D to 3D Simulation Universidad Veracruzana · 2610.09280 · ▲ 4
Thursday, October 8, 2026 · 71
- UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation Westlake University, Ant Group, Zhejiang University · 2610.09823 · ▲ 163
- Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight Salesforce, University of California San Diego, Texas A and M University · 2610.08077 · ▲ 162
- STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization Zhejiang University, Institute of Automation, Chinese Academy of Sciences, Tsinghua University · 2609.38169 · ▲ 120
- Long-WAM: Scaling the Context of World-Action Models NVIDIA, Massachusetts Institute of Technology, University of Hong Kong · 2610.10528 · ▲ 104
- nanoMuse: An Open-Source Personal Agent for Every Device You Own Zhejiang University · 2610.08699 · ▲ 101
- Recursive Game Creator: An Agentic Product-Level Experience-Oriented Game Harness University of Hong Kong, Shenzhen Loop Area Institute · 2610.08621 · ▲ 86
- GRACE: Generation-aware latent compression for efficient video generation KAIST, Kakao · 2610.10524 · ▲ 77
- DecepEval: A Benchmark for Evaluating Deception in LLM Agents Xi'an Jiaotong University, University of Virginia, Griffith University · 2610.07967 · ▲ 76
- Questioning the Questions: Sustaining Self-Evolution in Reasoning Models Washington University in St. Louis, University of Michigan, Ann Arbor · 2610.04299 · ▲ 71
- UniWAM: Unified World-Action Model arXiv · 2610.02054 · ▲ 61
- VepAgent: Bridging Causal-Transition via Tool-Augmented Reinforcement Learning for Video Event Prediction Nankai University, Carnegie Mellon University, Xiaohongshu · 2610.06293 · ▲ 57
- SGF+: Decoupling Gradient Flows for Autoregressive Video Generation Tsinghua University, JD.com, Chinese University of Hong Kong · 2610.10429 · ▲ 56
- Tetris3D: 3D Scene Generation With Objects That Fit Together KAIST · 2610.10539 · ▲ 45
- WorldSonus: Bringing Sound to Worlds Hong Kong University of Science and Technology, Noiz AI, MetaX · 2610.08760 · ▲ 35
- RunningTab: Direct Workspace Interaction with Environment-Side Tabs KAIST, DeepAuto AI · 2610.10444 · ▲ 35
- ReSAIL: Mitigating Collapse in Iterative Agent Self-Distillation Renmin University of China · 2609.39306 · ▲ 34
- RobotWorld: Benchmarking Multimodal Agents for Robot Use Across Diverse Tasks and Embodiments arXiv · 2610.10409 · ▲ 33
- Mechanics of Long-Context Hybrid Models Part 1.1: From Hybrid Attention to Hybrid Position Fudan University, Shanghai Innovation Institute, OpenMOSS Team · 2610.10114 · ▲ 32
- Semifactual Credit-Augmented Policy Optimization Zhejiang University, Westlake University, Shanghai Innovation Institute · 2609.40360 · ▲ 30
- Gains and Collapse in On-Policy Distillation:A Reinforcement Learning Perspective Zhejiang University, Westlake University · 2610.03185 · ▲ 30
- PhysEvo: Astra Can Act, Let It University of Chinese Academy of Sciences, Institute of Automation, Chinese Academy of Sciences, Tsinghua University · 2610.08995 · ▲ 30
- Recurrent Looped Transformer Princeton University, University of Pennsylvania · 2610.07591 · ▲ 28
- SWE-Game: Can Coding Agents Build the Games We Want? Shanghai Jiao Tong University, Zhongguancun Academy, Shanghai Innovation Institute · 2609.33678 · ▲ 26
- VIEScore2: Unified Image Evaluation with Spatially Grounded Explanations University of Waterloo, NVIDIA, National Taiwan University · 2610.00994 · ▲ 26
- MIMESIS: Learning User Simulators as Training Environments for Interactive Agents arXiv · 2610.09484 · ▲ 24
- Q-Learning with Scalar Adjoint Matching KAIST, RLWRLD · 2610.10437 · ▲ 23
- Inverting Multi-Vector Visual Document Indices Aalto University · 2610.09920 · ▲ 22
- NAMVIS: Next-Scale Autoregressive Multi-View Image Synthesis Applied AI Institute, T-Tech, King Abdullah University of Science and Technology · 2610.04722 · ▲ 19
- Agentic RAG Evaluation: Budget Allocation Across Questions, Trajectories, and Reads Carnegie Mellon University · 2610.05034 · ▲ 19
- On KL-Regularized Policy Optimization Princeton University · 2610.08963 · ▲ 19
- Mobile-4DGS: Unified Static-Dynamic Real-time Mobile Gaussian Splatting University of Technology Sydney, Yale University, City University of Macau · 2610.05289 · ▲ 18
- Minimal Witness Reinforcement Learning University of Pennsylvania, Shanghai AI Laboratory · 2610.07226 · ▲ 18
- On-Policy Distillation with Negative-Policy Rollouts NAVER AI Lab · 2610.07874 · ▲ 18
- AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video Generation Peking University, JD.com · 2610.10047 · ▲ 18
- WebFovea: When the Model Is Right but the Click Is Wrong -- Reliable Round Trips for Vision-Based Web Agents on Live Websites Independent researchers · 2610.03036 · ▲ 17
- Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning SperidLabs · 2610.09450 · ▲ 17
- From Pareto to Preference: Personalized Test-Time Scaling via Amortized Agentic Policy Discovery Beijing Institute of Technology, Xiaohongshu · 2610.09684 · ▲ 16
- CADFather: Autonomous CAD Reconstruction through Coordinated Tool Use Lomonosov Moscow State University, Innopolis University, FusionBrain Lab · 2610.09127 · ▲ 15
- A self-learning scientific agent for X-ray diffraction Hong Kong University of Science and Technology, Guangzhou, University College London, AI Lab of the Yangtze River Delta · 2610.07862 · ▲ 14
- DLoop: Looped Speculative Decoding NAVER AI Lab, NAVER AI Search Platform, Korea University · 2610.07659 · ▲ 13
- QuadTok: Quadtree Visual Tokenizer for Autoregressive Image Generation University of California San Diego · 2610.10497 · ▲ 12
- Inherit-MAS: Test-Time Evolution of Multi-Agent Systems through Workflow and Execution Inheritance University of Texas at Dallas, Independent researchers · 2610.02396 · ▲ 11
- Internalizing Agent Experience into Diffusion Model Weights via On-Policy Context Distillation Harbin Institute of Technology, Shanghai AI Laboratory, Shandong University · 2610.07250 · ▲ 11
- Learning Multimodal Embeddings with Evidence-Aligned Readout Hong Kong University of Science and Technology, Guangzhou, Tencent, Tsinghua University · 2609.33659 · ▲ 10
- SheetSage2: Coherent Lead-Sheet Transcription with Synthetic Supervision New York University, Mohamed bin Zayed University of Artificial Intelligence, ACE Studio · 2610.05336 · ▲ 9
- Improving Proactive AI Assistance with Hierarchical Procedural Understanding Sungkyunkwan University, Nanyang Technological University · 2610.06505 · ▲ 9
- SkillForge: Co-Evolving Skills and Agents via Dynamic Skill Lifecycles Institute of Computing Technology of the Chinese Academy of Sciences, University of California, Merced, Tsinghua University · 2610.09832 · ▲ 9
- Composing What Each Teacher Learned: Multi-Teacher On-Policy Distillation through Teacher-Relative Shifts Iowa State University, LinkedIn, Harvard University · 2610.10460 · ▲ 9
- Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute Corbenic AI · 2610.10845 · ▲ 9
- UniSkill: Learning Actor-Aligned Skill Proposals for an Evolving Policy Qiantang Credit, Ant Group, Chinese University of Hong Kong · 2610.10164 · ▲ 8
- EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory Hong Kong Polytechnic University, Hangzhou Diagens Biotechnology, University of Science and Technology of China · 2610.10533 · ▲ 7
- Rethinking World-Action Model for Compositional and In-Context Robotic Manipulation Peking University, AgiBot, CocoMatrix · 2610.02368 · ▲ 6
- StepCAD: Mesh-to-CAD Code Generation via LLM Policy and Geometry-Guided Search Massachusetts Institute of Technology · 2610.03799 · ▲ 5
- Agent Plasticity: Measuring Self-Improvement Through Experience University of California, Berkeley, Meta, University of Washington · 2610.08902 · ▲ 5
- RoboJEPA: Scaling Robotic Latent World Models Meta, Chandar Research Lab, Mila · 2610.10515 · ▲ 5
- Salt++: Context-Aligned Post-Training for Few-Step Streaming Multimodal Generation Hong Kong University of Science and Technology, Vivix Group Limited, University of Sydney · 2609.36995 · ▲ 4
- AutoResearch at Production Scale: Failure Modes and a Multi-Agent Framework Amazon · 2609.30541 · ▲ 3
- PAMI: Part Anchored Motion for Text to Human-Object Interaction Generation University of Tübingen, Max Planck Institute for Informatics · 2609.38466 · ▲ 3
- CoDance: Learning Reactive and Compliant Human-Humanoid Interaction from Video Duke University · 2610.05324 · ▲ 3
- DSReg: Provably Recovering Individual World Latents without Reconstruction University of Illinois Urbana-Champaign, Cold Spring Harbor Laboratory, Brown University · 2610.09457 · ▲ 3
- RoboQuest: Generalist Physical Agents that Search, Inspect and Test Nanyang Technological University · 2610.10388 · ▲ 3
- TIDES: Implicit Time-Awareness in Selective State Space Models ETH Zurich, Brown University · 2605.09742 · ▲ 2
- FastOPD: On-Policy Distillation for Lightweight VLA Deployment KAIST, Sungkyunkwan University · 2610.02832 · ▲ 2
- Beyond the Parameter Monolith: Reconstructive Memories, Executable Skills, and Residual Assembly for Language Models arXiv · 2610.04012 · ▲ 2
- How corner is a corner case? Percentile control for highway scenario generation University of Wisconsin-Madison · 2610.05003 · ▲ 2
- Lineage-Aware Memory Governance: A Derivation-Gated Framework for Privacy-Preserving Column-Level Access Control in Enterprise AI Agents Independent researchers, Sage Seven AI · 2610.07258 · ▲ 2
- Task-Sufficient Contraction: Source Selection for Machine Information Interfaces Ericsson Ireland · 2610.08884 · ▲ 2
- System Switch: When Should a Fast Decision Model Stop and Think? Independent researchers · 2610.09683 · ▲ 2
- LongTake: Learning to Sustain Dynamics in Long-Horizon Video Generation KAIST, Georgia Institute of Technology · 2609.38562 · ▲ 1
- ReGain: Restoring Subject Fidelity in Personalization on Synthetic Images University of Illinois Urbana-Champaign · 2609.38680 · ▲ 1
- Co-Evolving Robot Orchestrators and Policies through Deployment Stanford University, NVIDIA · 2610.09228 · ▲ 1
Wednesday, October 7, 2026 · 72
- Rethinking Cross-Tokenizer On-Policy Distillation: From Alignment Coverage to Supervision Reliability Alibaba · 2610.08448 · ▲ 201
- DuoMatching: Joint-Marginal Distribution Matching for Few-Step Video Generation Chinese University of Hong Kong, ByteDance, Hong Kong University of Science and Technology · 2610.03543 · ▲ 92
- TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models Alibaba, Ohio State University · 2610.07767 · ▲ 88
- CheckerBench: Can Long-Horizon Agents Synthesize Static-Analysis Checkers? East China Normal University, Humanlaya Data, Shanghai Jiao Tong University · 2610.07557 · ▲ 61
- EVISKILL: Grounding Skill Evolution in Replayable Evidence Jilin University · 2610.05030 · ▲ 52
- Kinematic MeanFlow: One-Step Action Generation Policy for Robotic Foundation Models Intel Labs China, Midea AI Research · 2610.00864 · ▲ 47
- From Evidence to Action: How Tool-Using Agents Fail National University of Singapore, Hong Kong Baptist University, Amazon · 2610.07753 · ▲ 45
- HuatuoGPT-3: RL-Only Domain Adaptation from Base Models Chinese University of Hong Kong, Shenzhen, Shenzhen Research Institute of Big Data, Shenzhen Loop Area Institute · 2610.05966 · ▲ 43
- AutoSciBench: Autonomous Benchmark Generation for Evaluating Scientific Agents Genentech, KAIST · 2610.05140 · ▲ 42
- TRIAGE: Direction-Aware Mismatch Stabilization of Native NVFP4 Reinforcement Learning InfiX AI, NVIDIA · 2610.07043 · ▲ 42
- Taming VLAs under Robot Execution Errors: Self-Compensation and Stress Testing POSTECH, Mohamed bin Zayed University of Artificial Intelligence · 2609.37334 · ▲ 40
- RLHND: Video Foundation Models as Physically Grounded Hand Trackers for Robot Learning RLWRLD, KAIST, Seoul National University · 2610.09455 · ▲ 37
- MiniCorp: The Last Mile of the AI Agent Firm Microsoft, University of Illinois Urbana-Champaign, Princeton University · 2610.05912 · ▲ 36
- World Action Learning via Interaction-Centric Spectral Latent Guidance Hong Kong University of Science and Technology · 2610.03607 · ▲ 35
- Selection-Based Structured Reasoning: Toward Efficient Multimodal Search Agents Apple, Carnegie Mellon University · 2610.01892 · ▲ 32
- Making LLMs Say What They Think: Measuring and Improving CoT-Interpretability Alignment New York University, NYU Shanghai · 2609.38972 · ▲ 27
- TimeBraid: Unifying Time Series and Language for Understanding and Forecasting University of California San Diego, University of Southern California, Aether AI · 2609.29792 · ▲ 26
- Learning Functional Subspaces for Neural Network Compression Helmholtz Munich, Technical University of Munich, Munich Center for Machine Learning · 2609.40127 · ▲ 25
- AGO AI Quality Gate: Evidence-First Release Decisions for Retrieval-Augmented Generation Protom Group · 2610.01218 · ▲ 25
- UNREAL: Unifying Retrieval and Long-Context with a Single Model NVIDIA, Technion · 2610.08463 · ▲ 24
- World Models' Last Exam in Physics Navers Lab at Einsia AI, Peking University, Tsinghua University · 2610.08791 · ▲ 24
- SlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving a2sys, KAIST · 2609.34117 · ▲ 23
- NeMo-DCR: Bit-Exact Delta-Compressed Refit for Scalable Agentic RL at Trillion-Parameter Scale Aalto University, NVIDIA · 2610.08430 · ▲ 22
- Rationale-Guided Policy Optimization: Learning to Reason with Adaptive Rationale Scaffolding New York University, University at Buffalo, Monash University · 2610.07342 · ▲ 21
- GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution Chinese University of Hong Kong, Shenzhen, Tianjin University, Harbin Institute of Technology, Shenzhen · 2610.00948 · ▲ 20
- DiffGate: Difficulty-Gated Teacher Guidance for On-Policy Distillation Indian Institute of Science, Snap · 2610.04596 · ▲ 20
- EmbodiedSmith: Scaling Embodied Data through Recursive Self-Improvement Flywheel in Simulation Hong Kong University of Science and Technology, Guangzhou, Xiaomi, Southeastern University · 2610.07969 · ▲ 20
- Adaptive Latent Capacity for World Models Arm AI Research at Arm · 2609.32921 · ▲ 19
- OPD Before RL: Warm-Starting Rubric-Based RL with On-Policy Distillation New York University, Meta · 2610.02781 · ▲ 16
- DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling University of Wisconsin-Madison, Microsoft, Nanyang Technological University · 2610.04933 · ▲ 15
- HiPLEX: Hierarchical Policy Factorization for Full Duplex Speech Language Models Qualcomm AI Research, KAIST · 2610.07727 · ▲ 15
- AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World Model Mohamed bin Zayed University of Artificial Intelligence, Amazon, Massachusetts Institute of Technology · 2610.08773 · ▲ 15
- Building Rome from a Single Image Applied Intuition, Purdue University, University of Illinois Urbana-Champaign · 2610.08790 · ▲ 15
- Understanding and Enhancing Backdoor Persistency in LLM Agent Post-Training University of Illinois Urbana-Champaign, MATS Research, Independent researchers · 2610.07510 · ▲ 14
- Hiding Tool Latency in On-Device Cascaded Voice Agent through Speculative Execution Qualcomm AI Research, KAIST · 2610.07641 · ▲ 14
- Sherpa: Teaching LLMs to Teach Adaptively Stanford University, Georgia Institute of Technology, Carnegie Mellon University · 2610.08778 · ▲ 14
- Harness-Aware Distillation for Small Language Model Agents KAIST, Independent researchers · 2610.02858 · ▲ 13
- We Query, Therefore We Compute: On Oracle Computation beyond the Machine, with an Application to Agents Institute of Computing Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Nanjing University · 2610.09243 · ▲ 13
- Personal-Agent Mediated Recommendation with Cross-Platform User History University of California San Diego, Meta · 2610.07588 · ▲ 12
- Towards In-Parameter Memory Augmentation for Large Language Models Hong Kong University of Science and Technology, Hong Kong Baptist University · 2610.08630 · ▲ 12
- Harness Engineering for Software Engineering via Modular Executable Dev-Primitives University of Illinois Urbana-Champaign, Pennsylvania State University · 2610.07832 · ▲ 11
- HLA: Expressive Hybrid Linear Attention via Chunk-Wise Dynamic Mixing Monash University, Zhejiang University · 2610.05842 · ▲ 10
- DAEDALUS: Bootstrapping Agent Memory from Self-Generated Tasks Illuin Technology · 2610.08048 · ▲ 10
- ALIVE: Interaction-Aligned Object Insertion for First-Frame-Guided Video Editing University of Rochester, Adobe · 2610.08779 · ▲ 10
- Multilinguality in Hybrid Attention LLMs University of California, Los Angeles, Fudan University · 2609.35378 · ▲ 9
- DistScene: Object-to-Scene Distillation for 3D Scene Generation Hong Kong University of Science and Technology, TeleAI, China Telecom · 2610.06960 · ▲ 9
- Learning to Read the Contextual Tokens in Diffusion Transformers Tel Aviv University, Cornell University · 2610.06844 · ▲ 8
- Accent Analogy Guidance: More Speaker Similarity at Equal Accent in Cross-Lingual Voice Cloning ESTsoft · 2609.29123 · ▲ 7
- Attacca: Goal-Directed Control under State Continuity for Long-Horizon Embodied Agents Nanyang Technological University · 2610.07785 · ▲ 6
- VeriFine: Scaling Verification for Self-Improvement in Embodied Reasoning NVIDIA, University of California, Los Angeles, University of California, Berkeley · 2610.08761 · ▲ 6
- Do Language Models Need a Trainable Input Embedding Table? Fixed Minimal Token Codes at 1.7B-Class Scale arXiv · 2610.04002 · ▲ 4
- Learning Discriminative Geometry for Drifting Models ELLIS Institute Finland, Aalto University · 2610.04703 · ▲ 4
- JLD: Perceptual Distance Through A Jacobian Lens University of Texas at Austin, Google, University of Colorado Boulder · 2610.05967 · ▲ 4
- Judged Useless, Queried Anyway: Tool-Using Agents Rarely Turn Their Own Evidence Judgments into Stopping Decisions Nanyang Technological University, National University of Singapore, Agency for Science, Technology and Research · 2610.06191 · ▲ 4
- Stepped MoE: Segment-Level Routing with Configurable Inference Complexity Apple · 2610.07348 · ▲ 4
- Sensor-Language-Action Models University of California, Los Angeles · 2610.08244 · ▲ 4
- CtrlCache: Accelerating Interactive Video World Models with Control-Aware Caching University of Auckland, University of New South Wales, Sydney · 2610.08777 · ▲ 4
- DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation Texas A and M University, ByteDance · 2610.02188 · ▲ 3
- A Safe Action Is Not Enough: Feasible-Future Decoding for Vision-Language-Action Policies Huawei, Technical University of Berlin, UCL Center for AI · 2610.05166 · ▲ 3
- MEND: RL For Flow Models via Proximal Velocity Matching University of Texas at Austin, Google, University of Colorado Boulder · 2610.05954 · ▲ 3
- JumpStart Your Policy Learning with Lessons from 160,000 Training Runs University of Washington, Microsoft · 2609.13730 · ▲ 2
- Toward Real-Time VLAs: Stage-Aware Two-Step Flow Denoising and System-Level Evaluation Magiclab Robotics, Zhejiang University, Southeast University · 2609.39822 · ▲ 2
- Magic-W0: A Structured World-Action Foundation Model for Physical Intelligence Magiclab Robotics · 2609.39870 · ▲ 2
- Source Identification Is Not Fitness Testing: Measuring the Limits of Synthetic-Data Attribution Ericsson · 2610.00417 · ▲ 2
- Cross-Lingual Alignment for Decoder-Only Models using MoE Routers University of California, Los Angeles, Haverford College, Mila · 2610.01921 · ▲ 2
- ConEx: Human-Interpretable Saliency Maps via Concept-Aware Attribution Tel Aviv University, Open University · 2610.04605 · ▲ 2
- Technical Report on the Turba Fertilizer Machine Learning Stack in Morocco Turba Open Lab · 2610.05949 · ▲ 2
- Execution-Aligned Progressive Noise for Consistent Asynchronous Replanning in Generative Robot Policies Magiclab Robotics, Southeast University · 2610.06090 · ▲ 2
- Conditional Trajectory Peaks: Single-Pass Multimodal Policies over Action Chunks Magiclab Robotics, Southeast University · 2610.06104 · ▲ 2
- S2PD: Serial-to-Parallel Diffusion for Physically and Logically Consistent Video Generation University of Cambridge, Toyota Motor Europe · 2610.06847 · ▲ 2
- Structuring MoE Expert Selection for Agentic Reinforcement Learning Apple, Purdue University · 2610.07332 · ▲ 2
- WildMatch: Weakly Supervised Image Matcher Adaptation for Wildlife Re-Identification Jagiellonian University, Jagiellonian Center for Artificial Intelligence, NASK National Research Institute · 2610.07384 · ▲ 2
Tuesday, October 6, 2026 · 81
- Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation Kandinsky Lab · 2610.05608 · ▲ 161
- LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient Sketches NVIDIA · 2610.06647 · ▲ 84
- ALoDLM: Adaptively Looped Diffusion Language Models University of Illinois Chicago, Amazon, Korea University · 2610.04198 · ▲ 74
- Memadapter: Counterfactual Adaptation Against Memory-induced Sycophancy Jilin University, Xiamen University · 2610.05162 · ▲ 59
- In-Distribution Forcing for Long Video Generation at Test Time Seoul National University, Korea University, Sungkyunkwan University · 2610.03120 · ▲ 49
- Foundations of Proactive Agents: Principles, Technical Layers, and Proactivity-Gym KAIST, University of Minnesota · 2609.37267 · ▲ 40
- LMBuild: Evaluating LLM Agents for Generating Buildable and Functional Structures University of Illinois Urbana-Champaign · 2610.04292 · ▲ 38
- SEER: Self-Evolving Event Reasoning and Retrieval for Time Series Forecasting Google, University of Virginia · 2610.04109 · ▲ 32
- When to Switch: Reliable Action-Chunk Extension for Vision-Language-Action Models KAIST, Gwangju Institute of Science and Technology, NVIDIA · 2610.05719 · ▲ 32
- Representation-Space MMD for Diffusion Language Models Yandex Research, HSE University, Applied AI Institute · 2610.06648 · ▲ 31
- Towards Looped Models Done Right, Part II: Rethinking at Fixed Points Institute of Foundation Models, University of Southern California, Carnegie Mellon University · 2610.06833 · ▲ 30
- World Editing: Intervening on Executable Worlds at Increasing Depth University of Waterloo, G-G-G, Comfy Org Research · 2610.02331 · ▲ 29
- OSWorld-Pro: Process-based Evaluation for Computer Use Agents NVIDIA · 2609.24890 · ▲ 28
- Data Unlearning via Inverse Distillation AI Foundation lab, MIRAI, Applied AI Institute · 2609.36099 · ▲ 27
- OmniReasoning: Pushing the Limits of Audio-Visual Joint Reasoning Peking University, Alibaba · 2609.39490 · ▲ 27
- Self-Generated Feedback Destabilizes Test-Time Training: A Causal Decomposition of Long-Horizon Adaptation KAUST · 2610.05076 · ▲ 25
- SearchJev: A Fast and Calibrated System-1 Model for Search Agents Huawei, Leiden University · 2610.05107 · ▲ 24
- Optimizing the Optimizer: Language Models Discover Faster Molecular Relaxation AXXX, ITMO University, Zelinsky Institute of Organic Chemistry of the Russian Academy of Sciences · 2610.06577 · ▲ 24
- CANOPY: Adaptive-Granularity Evidence Compression for Multimodal RAG POSTECH · 2610.00923 · ▲ 23
- ASCENT: Online Test-Time Training of Long-Horizon Agents via Self-Distillation of Verified Experience UNSW Sydney · 2610.05303 · ▲ 23
- Base Models Can Reason By Taking a Cue From Training Data Massachusetts Institute of Technology, University of California, Berkeley, University of Washington · 2610.06851 · ▲ 23
- UndoBench: Separating Task Competence from Recovery Capability in Tool-Using AI Agents Independent researchers · 2610.05622 · ▲ 21
- ProgressCompass: Embodied Progress Reward Models Are Lost Without the Right Context Northwestern University, University of California, Santa Barbara, University of Illinois Urbana-Champaign · 2609.36684 · ▲ 20
- HLA-WM: Hybrid Linear Attention for Long-Horizon Video World Models Monash University, University of Adelaide, Zhejiang University · 2610.05739 · ▲ 20
- RobotUse: Allocating Computation, Context, and Decisions KAIST, Seoul National University · 2610.04929 · ▲ 19
- RealtimeWAM: One-Step Asynchronous World Action Models Nanyang Technological University, Beihang University, SenseTime · 2610.06617 · ▲ 19
- QuantCode Model: Specializing Language Models for Executable Algorithmic Trading Code arXiv · 2609.39420 · ▲ 18
- PerturBot: Breaking Shortcut Priors in Vision-Language-Action Models with Perturbative Training Zhejiang University, Shanghai Innovation Institute, University of Hong Kong · 2610.04616 · ▲ 18
- Beyond Semantic Similarity: Performance and Costs of Agentic Retrieval for Complex Tasks NVIDIA, University of Edinburgh · 2610.05750 · ▲ 18
- Capability-Driven Self-Evolution of Agent Memory University of Science and Technology of China, Microsoft, University of California San Diego · 2610.06361 · ▲ 18
- Rethinking Long-Video Efficiency: A Joint Allocation Perspective on Frames, Pixels, and Front-End Latency University of Central Florida, Meta, Axon · 2610.04318 · ▲ 17
- Prism: Dynamic Sparse Attention for Native 2K Joint Video-Audio Generation Model Training Fudan University, Tencent, Zhejiang University · 2610.05416 · ▲ 17
- When Does Selection Replace Extraction? A Pre-Registered Test of Agent Memory with a Typed Decision Model Independent researchers · 2609.34227 · ▲ 16
- Video2Skill: From Streaming Experience to Reusable Embodied Skills Northwestern University, Carnegie Mellon University, University of Illinois Urbana-Champaign · 2609.36691 · ▲ 16
- DeskForge: Dense Supervision from Desktop Environments for Computer-Use Agents ETH Zurich, IBM, Microsoft · 2610.02320 · ▲ 16
- Certification of Real Images through Calibrated Content Authentication Mohamed bin Zayed University of Artificial Intelligence · 2610.05870 · ▲ 16
- Noise Out, Bias In: Targeted Bias Injection in Diffusion Language Models via Closed-Loop Activation Steering Mohamed bin Zayed University of Artificial Intelligence · 2610.05894 · ▲ 16
- The Missing Primitive: Diagnosing and Repairing Mathematical Reasoning in Large Language Models Texas A and M University, Harvard University, Vanderbilt University · 2610.02191 · ▲ 14
- How to Loop MoE: Flatten the Experts, Untie the Attention Case Western Reserve University, Kyoto University, National Institute of Informatics · 2609.35751 · ▲ 13
- LLM-as-Jev: LLMs Are Already Jev-Style Decision Models -- When and How to Fine-Tune Them arXiv · 2610.02076 · ▲ 13
- What Matters for Latent Reasoning with Flow Matching Samsung, Technical University of Iasi, Queen Mary University of London · 2610.06666 · ▲ 13
- From Knowledge Access to Source Learning: Developing Source-Specific Competence Georgia Institute of Technology, University of Illinois Urbana-Champaign, University of California, Los Angeles · 2610.02150 · ▲ 12
- What Gradients Add to Text Leakage in Split Language Models, Counted per Token and per Document Setloop.io · 2610.04128 · ▲ 12
- Dynamic Harness Search: Building Multi-Agent Systems Per-Query via Prediction Google, Arizona State University · 2610.04137 · ▲ 12
- LiFT: Loop Flow Transformers University of Amsterdam, TNO · 2610.05538 · ▲ 12
- Empirical Variational Autoencoder University of Tokyo · 2610.06545 · ▲ 11
- PaLoRA: Paced Low-Rank Adaptation for Continual Learning Peking University, Institute of Automation, Chinese Academy of Sciences · 2610.04226 · ▲ 10
- TextReg: Mitigating Prompt Distributional Overfitting via Regularized Text-Space Optimization Georgia Institute of Technology, University of Illinois Urbana-Champaign · 2605.21318 · ▲ 9
- Training Numerical Intelligence via Auto-Diagnosis and Skill Discovery University of California, Berkeley, Alibaba · 2610.03872 · ▲ 9
- Periscope: Extending Frozen Language Models Beyond Their Context Window KAUST, Edge Hill University · 2610.04047 · ▲ 9
- PluginRSI: Recursive Improvement of Agent Harnesses with Reusable Plugins University of Science and Technology of China, Meituan, Wuhan University · 2609.32423 · ▲ 8
- MM-ABC: Towards Generalist Mobile Manipulation via Seeing, Coordinating and Imagining Xspark AI, Hong Kong University of Science and Technology, Guangzhou, Tsinghua University · 2609.35652 · ▲ 7
- GeoSET: Generalist Foundation Model for SAR-to-EO Image Translation KAIST · 2609.37496 · ▲ 7
- MEA: A Reward-Driven Multi-Agent System for Faithful Model Explanations University of Virginia, Vellore Institute of Technology, Indian Institute of Technology Patna · 2610.02480 · ▲ 7
- Labels Override Definitions in Jev-Style Typed Decision Models University of Southern California · 2610.02586 · ▲ 7
- OmniConfess: Eliciting Token Confessions to Mitigate Omni-Modal Hallucination Beijing University of Posts and Telecommunications, Nanyang Technological University · 2610.02999 · ▲ 7
- COSMI: COmpositional Synthesis of Multi-object Interactions University of Tübingen, Tübingen AI Center, Zuse School ELIZA · 2610.03252 · ▲ 7
- The Numerical Linear Algebra of Large Language Models Qatar Computing Research Institute, Hamad Bin Khalifa University, University of Minnesota · 2610.04631 · ▲ 7
- Code2Games: Enabling Coding Agents for Gaming World Generation Peking University, La Trobe University · 2610.05033 · ▲ 7
- Learning to Learn a Language Fraunhofer Institute for Algorithms and Scientific Computing, University Hospital Bonn · 2610.05879 · ▲ 7
- Collaborative Personalized Preference Alignment for LLMs under Data Deficiency City University of Hong Kong, University of Washington, Hong Kong University of Science and Technology · 2610.05898 · ▲ 7
- SoK: Semantic Decision Engines in Network Control Loops University of Technology Sydney · 2610.06425 · ▲ 7
- Intent Interpretation at RIC Timescales: Jev Decision Models versus Large Language Models in 6G Open RAN University of Technology Sydney · 2609.23136 · ▲ 6
- Learning Steadily: Accumulating Relative Point Margin Scores for Face Image Quality Assessment Fraunhofer Institute for Computer Graphics Research, Technical University of Darmstadt · 2609.31662 · ▲ 6
- OpenRUA: Robot-Use Agents Are Zero-Shot Visuomotor Policies University College London, Carnegie Mellon University · 2610.02459 · ▲ 6
- Learning Latent Protein Languages for Autoregressive Generation University of Missouri, University of Southern California, Independent researchers · 2610.03978 · ▲ 6
- UnAct: Gradient-Free Unlearning via Targeted Activation Intervention National Institute of Technology Srinagar · 2610.04426 · ▲ 6
- FairRSFM: A Biome-Aware Benchmark and Debiasing Framework for Remote Sensing Foundation Models Space Applications Centre, ISRO, Indian Institute of Technology Bombay, Indian Institute of Science Education and Research Bhopal · 2610.05790 · ▲ 6
- Closing the Context Gap: Activation Alignment for Tabular In-Context Learning Independent researchers · 2610.06679 · ▲ 6
- Sharpen Without Search: On-Policy Distillation of Sequence-Level Power Distribution University of Southern California, Intel AI · 2610.06804 · ▲ 6
- Adaptive Fused Prior Transfer for Controllable Generative Image Compression Santa Clara University · 2605.16817 · ▲ 5
- GeoCR: Learning a Generalist Cloud Removal Prior from Heterogeneous Observations KAIST · 2609.32510 · ▲ 5
- iADD: Improving Alignment and Diversity in Diffusion Policy Optimization Tribhuvan University, NAAMII, INSAIT, Sofia University · 2610.01789 · ▲ 5
- The AI Theorist reveals excitonic structure in α-RuCl_3 University of Oxford, University of Waterloo, Stanford University · 2610.02417 · ▲ 5
- CurveCodec 2: Skeleton-agnostic animation compression with a learned entropy model University of Hong Kong, Adobe · 2610.04211 · ▲ 5
- Agentic discovery of blood biomarker from distilled private health records Harvard Medical School, Clalit Research Institute, Ben-Gurion University of the Negev · 2610.04749 · ▲ 5
- Arm-wise Compositional Generalization in Dual-Arm Vision-Language-Action Models Dalian University of Technology, University of Oxford, Shanghai Jiao Tong University · 2610.06184 · ▲ 5
- Adapting prior-data fitted networks for tabular anomaly detection Technion, Israel Institute of Technology · 2610.06693 · ▲ 5
- InterMimicGen: Scaling Humanoid Loco-Manipulation through Self-Evolving Motion Imitation University of Illinois Urbana-Champaign, NVIDIA · 2610.06850 · ▲ 5
- Foresight: planning future perception in streaming VLMs without retraining Independent researchers, NAAMII, University College London · 2610.03123 · ▲ 4
- DEPICT: Scoring Text-to-Image Alignment by Answer Agreement Sword Health, NOVA University of Lisbon · 2610.03617 · ▲ 4
Monday, October 5, 2026 · 65
- RealCompanion: Benchmarking Human Understanding from Reasoning over Longitudinal Real-World Conversations Quis Lab, Independent researchers · 2610.01780 · ▲ 275
- Does Learning Protein Folding Generalize to Broader Reasoning? Shanghai Jiao Tong University, Fudan University, Shanghai Innovation Institute · 2609.38879 · ▲ 120
- FrameMorrow: Future-guided Frame Selection with Prospective Tokens for Long-Horizon Video Generation National University of Singapore, Harbin Institute of Technology, Shenzhen · 2609.38839 · ▲ 112
- MotorMind: Scaffolding General Vision Language Models for Zero-Shot Robot Manipulation University of Illinois Urbana-Champaign · 2609.38078 · ▲ 111
- Scaling Trajectories for Complex Tasks through Recursive Self-Rewrite Tencent, University of Maryland, University of Georgia · 2610.02826 · ▲ 103
- World Action Modeling with Progressive Visual Planning Shanghai Jiao Tong University, Shanghai Innovation Institute, Meta · 2610.02508 · ▲ 96
- Native Action-Prior Learning from Videos for World Action Models Meta, University of Copenhagen, Imperial College London · 2610.03391 · ▲ 88
- On-Policy Parameter Update Direction Underlies Generalization in LLM Post-Training Institute of Computing Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Meituan · 2609.36659 · ▲ 87
- Latent-MOPD: Latent Multi-Teacher On-Policy Distillation Case Western Reserve University, Zillow Group, University of Illinois Chicago · 2610.02381 · ▲ 70
- LexReward: A Taxonomy-Driven Reward Framework for Legal Language Models Peking University, Northeastern University, Tsinghua University · 2609.39071 · ▲ 64
- PDE-JEPA: Predictive Representation Learning of Latent Dynamics Modeling for Parametric PDEs Zhejiang University · 2609.34715 · ▲ 62
- Pivot-SD: Efficient Self-Distillation for Masked Diffusion Language Models KAIST, University of Toronto, Vector Institute · 2610.03665 · ▲ 62
- PointWAM: 3D World Action Modeling for Dexterous Robotic Manipulation POSTECH, KAIST, Hanyang University · 2610.02840 · ▲ 61
- HyperBrowseComp: A Multilingual and Multimodal Stress Test for Web-Browsing Agents Mohamed bin Zayed University of Artificial Intelligence, Mila, the Quebec Artificial Intelligence Institute, Inception AI · 2610.03574 · ▲ 60
- Science or Slop?: Benchmarking and Mitigating Scientific Slop in AI-Generated Papers Seoul National University, University of Minnesota · 2610.00531 · ▲ 59
- VeriHarness: Scaling Agentic Verification for Long-Horizon Tasks University of Cambridge, Google · 2610.00972 · ▲ 58
- EditHero: A Benchmark for Long-Horizon Part-Level 3D Editing and Vibe Modeling Alaya Lab, University of Tokyo, Institute of Science Tokyo · 2610.02298 · ▲ 58
- Spatial Memory Intelligence: Endowing World Models with Understanding-Driven Long-Term Memory Chinese University of Hong Kong, Shenzhen, Alibaba, Shenzhen Loop Area Institute · 2610.02521 · ▲ 57
- SimuVerity: Benchmarking Agents for Engineering-Grade Simulink Model Generation Xi'an Jiaotong University · 2610.02304 · ▲ 54
- World Embedding Benchmark arXiv · 2610.03632 · ▲ 53
- Multilingual GSM-Symbolic: What determines capability transfer across languages? Aarhus University, Danish Foundation Models, University of Alabama · 2610.03367 · ▲ 51
- Science Utopia? Closed-Loop LLM Simulation of Academic Research Ecosystems Georgia Institute of Technology, University of California, Los Angeles, University of Chicago · 2610.01257 · ▲ 45
- Source Preference in the Wild: How LLM Agents Favor Items by Source, and How to Reduce It Seoul National University · 2610.03195 · ▲ 45
- Triadic Linear Attention: Three-Dimensional Recurrent States for Long-Context Sequence Modeling Massachusetts Institute of Technology, IBM · 2609.36529 · ▲ 37
- HelixWorld: A Real-time Interactive Audio-Visual World Model Hong Kong University of Science and Technology, Noiz AI · 2609.38123 · ▲ 37
- Language Models that Play Chess and Explain Their Moves Princeton University · 2610.03695 · ▲ 37
- DEFINE: Exemplar-Guided Accent Control for Zero-Shot TTS Ca' Foscari University of Venice, Kandinsky Lab, Sleeping AI · 2609.32777 · ▲ 34
- MetaRubric: Learning to Reward for Rubric-Based Reinforcement Learning Nanyang Technological University, Singapore · 2610.02824 · ▲ 33
- ProAR: Learning Prospective Reasoning with Autoregressive Video Models Hong Kong Polytechnic University, University of California, Davis, Microsoft · 2610.03664 · ▲ 32
- 4DCodeBench: Benchmarking Agents on Inverse Graphics of Dynamic Scenes Johns Hopkins University, Stanford University, Max Planck Institute for Intelligent Systems · 2610.03715 · ▲ 30
- Learning from Runtime Feedback through Failure-Bank Self-Evolution for Vision-Language-Action Models University of Notre Dame · 2609.39820 · ▲ 29
- Local Support Learning Tel Aviv University, Massachusetts Institute of Technology · 2610.02126 · ▲ 28
- FrugalEvo: Towards Cost-Aware LLM-Guided Program Evolution National University of Singapore, University of Washington, Princeton University · 2610.03675 · ▲ 28
- Beyond Future Prediction: Denoising as Generative Adaptation for Robot Control University of California San Diego, University of Hong Kong, Hong Kong University of Science and Technology · 2609.28339 · ▲ 27
- WEFT: Scaling Tool-Use Post-Training for General-Purpose Agents East China Normal University, Fudan University, Shanghai Innovation Institute · 2609.36887 · ▲ 24
- Equal Ranking Quality, Different Decisions: Measuring and Reducing Order Dependence in LLM Scorers Thomson Reuters Labs, University of Toronto, Vector Institute · 2608.26762 · ▲ 23
- Rollout-Marginal Distillation for Long-Horizon Autoregressive Video Generation Chinese University of Hong Kong, Tencent · 2609.37925 · ▲ 23
- Looping Beyond Twice: A Scalable Recipe for Looped Mixture-of-Experts Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Peng Cheng Laboratory, University of Chinese Academy of Sciences · 2610.01153 · ▲ 23
- Where Does Retrieval-Based Open-Ended Evaluation Fail? Automatic Taxonomy Induction from Long-Form Medical Answer Factuality Verification Johns Hopkins University · 2609.30467 · ▲ 22
- Diptych: Scoped, AI-Interpreted Comparison for Reference Listening in Music Production Zhejiang University, Singapore University of Technology and Design · 2609.39963 · ▲ 21
- From Gradients to Capabilities: Understanding Multi-Teacher On-Policy Distillation University of Illinois Urbana-Champaign, Princeton University, Westlake University · 2610.02179 · ▲ 20
- Collective Bias Mitigation via Model Routing and Collaboration Nanyang Technological University, National University of Singapore, Harbin Institute of Technology · 2610.03240 · ▲ 20
- LVMT: Video Mask Transformer for Long-term Video Segmentation Eindhoven University of Technology, RWTH Aachen University · 2609.34895 · ▲ 19
- Tail-Influence Sampling for CVaR Policy Evaluation Imperial College London, Huawei, UCL Centre for AI · 2609.38096 · ▲ 19
- QuantWM: Temporally Consistent 2-Bit KV Cache Quantization for Video World Models Harbin Institute of Technology, Shenzhen, National University of Singapore · 2609.26425 · ▲ 18
- Covert Assistance: Helpful LLM Agents Evade Oversight in Multi-Agent Systems University of Illinois Urbana-Champaign, Genies, NVIDIA · 2609.39050 · ▲ 18
- Skill2Real: Agentic Skill Learning for Zero-Shot Sim-to-Real Robot Manipulation University of California, Los Angeles, Shanghai Jiao Tong University, Envora · 2610.02788 · ▲ 18
- Efficient Reasoning Training Does Not Always Harm CoT Faithfulness and Monitorability University of Sheffield, AstraZeneca · 2610.03509 · ▲ 17
- MinkowskiPE: Minkowski Positional Encoding for Spatiotemporal Perception Wave Intelligence Lab, Chinese University of Hong Kong, Shanghai AI Laboratory · 2609.33804 · ▲ 16
- Octrees as an Explicit 3D Language Peking University, Independent researchers · 2610.02388 · ▲ 16
- GTR: Gated Token Recurrence for Efficient Dense Prediction Didi International Business Group, Intellindust AI Lab, Institute of Automation, Chinese Academy of Sciences · 2609.26590 · ▲ 15
- Self-Supervised Scaling of Terminal Environments for Scientific Domains Tencent, University of Georgia, University of Maryland · 2610.02710 · ▲ 13
- Receiver-Conditioned Latent Communication gives 94% CacheBack Columbia University, Amazon · 2609.32046 · ▲ 12
- Rethinking Token Reweighting for SFT: Suppress, Reverse, and Extrapolate Learned Features Shanghai AI Laboratory, University of Science and Technology of China, Fudan University · 2609.33463 · ▲ 12
- WM-VLM: Probing Internal World Models for Interleaved Visual-Textual Reasoning University of California San Diego, Mohamed bin Zayed University of Artificial Intelligence, Carnegie Mellon University · 2609.34826 · ▲ 12
- From Retrieval to Typed Decisions: Calibrated System One Models from Biomedical Sentence Encoders Queen's University Belfast · 2610.02486 · ▲ 12
- EyeRobot 2.0: Active Gaze for Precise Manipulation without Wrist Cameras University of California, Berkeley, Amazon, Stanford University · 2610.03710 · ▲ 12
- KeyRec: Bounded Visual Memory for Streaming and Long-Video Understanding University of Virginia, Google · 2609.32182 · ▲ 11
- Dream4ACT: A Shared Visual Action Interface for Multi-Embodiment Video-Action Modeling Beijing Humanoid Robot Innovation Center, Beijing Institute of Technology, Harbin Institute of Technology, Shenzhen · 2609.40153 · ▲ 10
- Can Computation from Earlier Problems Help LLMs Solve New Ones? Renmin University of China, University of Modena and Reggio Emilia · 2609.39394 · ▲ 9
- Improving Atomic-Fact Recall via Focused Views in Unstructured Knowledge Editing Georgia Institute of Technology, University of Pennsylvania, University at Albany · 2610.02772 · ▲ 8
- The Extender: A Log-Structured Transformer University of Illinois Chicago · 2609.32759 · ▲ 7
- Strike a Chord! Modal Kinetic Typography Simon Fraser University, Adobe · 2609.38325 · ▲ 7
- CUAWright: A Minimal Unified Interface for Digital Agents Microsoft, National University of Singapore, Ohio State University · 2610.04116 · ▲ 6
- JEPA-TTT: Persistent Test-Time Training of Latent World Models for Planning under Dynamics Shifts Honda Research Institute USA, Johns Hopkins University · 2610.00722 · ▲ 3
Friday, October 2, 2026 · 84
- OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction Nanjing University, Shanghai AI Laboratory, JD.com · 2610.01762 · ▲ 223
- On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics University of Cambridge · 2609.35259 · ▲ 185
- GraphForge: Training Working Agents with Graph-Anchored Workspace Synthesis University of Science and Technology of China, Fudan University, Shanghai Innovation Institute · 2609.38923 · ▲ 146
- Adaptive Reward Routing: Dynamic Multi-Reward Optimization for Joint Audio-Video Diffusion via Forward-Process RL Hong Kong University of Science and Technology, Tencent, University of Hong Kong · 2609.37200 · ▲ 128
- Predictive Credit: Measuring What Scientific Explanations Add to Experimental Forecasts Carnegie Mellon University, Tsinghua University · 2610.00314 · ▲ 98
- Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States Nankai University, Alibaba, Tsinghua University · 2610.01415 · ▲ 86
- Sharpening Tax in Post-Training Meta, University of Wisconsin-Madison, New York University · 2610.01509 · ▲ 86
- Prefill-Free Cross-Family KV Cache Transfer for Heterogeneous Multi-Agent LLMs University of Southern California, Seoul National University, New York University · 2609.32259 · ▲ 84
- Hierarchical Continuous Diffusion Language Models University of Illinois Urbana-Champaign, Amazon · 2610.02193 · ▲ 82
- Agent Priors-guided Policy Learning National University of Singapore · 2609.35690 · ▲ 79
- World Observer: Joint Actor-Observer Generation for Persistent World Modeling KAIST · 2610.02162 · ▲ 79
- Transformers Stop Thinking Too Early, and a Tiny LoRA Fixes It Georgia Institute of Technology · 2609.36585 · ▲ 73
- A Missing Piece for Trustworthy AI Reviewers: From Benchmarking Rhetorical Robustness to SciCore Review Virginia Tech, University of Maryland, Mohamed bin Zayed University of Artificial Intelligence · 2609.39027 · ▲ 73
- ActiveSaddler: Automated Curriculum Learning for Agent Harness Optimization POSTECH, KAIST, Microsoft · 2610.00906 · ▲ 71
- X-Tree: Tokenizing Reusable Experience for Efficient Agent Generalization University of Waterloo, Duke University, National University of Singapore · 2609.32993 · ▲ 70
- ROWBench: Do Video Models Render What the Program Specifies? Alaya Lab · 2610.02205 · ▲ 66
- E-MoE: Enhanced Mixture-of-Experts for Non-Factorized Diffusion Language Models Applied AI Institute · 2609.37533 · ▲ 62
- Retrieval-Augmented Skill Optimization via Cross-Harness Adaptation Korea University, KAIST, Meta · 2609.38024 · ▲ 60
- EgoTools: Towards Tool-Centric Reasoning in Real-World Egocentric Videos Nanyang Technological University, A-STAR, KAIST · 2609.39378 · ▲ 60
- Make Sparse Rewards Count: Density-Aware Reward Aggregation for Multi-Reward RL University of Chinese Academy of Sciences, University of Minnesota Twin Cities, University of Wisconsin-Madison · 2610.00574 · ▲ 60
- Video Generation Models: A Survey of Post-Training and Alignment Arizona State University, Twitch, Stanford University · 2610.00812 · ▲ 57
- InterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-Manipulation University of Illinois Urbana-Champaign · 2610.02196 · ▲ 57
- AutoGUIWorld: Image Generators as Visual World Models for GUI Agent Tencent · 2610.01215 · ▲ 56
- Decentralized Master-Mind: Joint Action Refinement through Iterative Intent Denoising in Multi-Agent Pathfinding CogAI Lab · 2609.32019 · ▲ 53
- Persona Dosing: Calibrated Activation Steering for Graded Trait Control Georgia Institute of Technology, Zhejiang University, University of British Columbia · 2609.36388 · ▲ 51
- Scaling and Distilling Text Embeddings for Better Diffusibility University of Michigan · 2610.01016 · ▲ 51
- Beyond the Current Scene: Event-Referential Grasping with Active View Selection Seoul National University, NVIDIA · 2609.39375 · ▲ 50
- Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens Peking University, National University of Singapore, NVIDIA · 2610.01939 · ▲ 47
- Architect-Ant: Editable Automatic Furnishing of Architectural Floor Plans King Abdullah University of Science and Technology, Miami University · 2606.10953 · ▲ 45
- Decoding Looped Transformers Better for (Almost) Free Apple · 2610.02185 · ▲ 43
- AutoDataBench: A Data-centric Testbed for Accelerating Auto Research Hong Kong Polytechnic University, IQuest Research, Shanghai Jiao Tong University · 2609.40097 · ▲ 40
- 4Director: Controlling Video World Models with Rigid 3D Geometry Stability AI, University of Illinois Urbana-Champaign · 2610.02160 · ▲ 36
- SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation University of Illinois Urbana-Champaign · 2610.02201 · ▲ 36
- CorrGRPO: Correlation-Normalized GRPO for Multi-Reward Learning Hong Kong University of Science and Technology · 2609.36820 · ▲ 32
- RPTune: Learned Context Curation for LLM Catalog Search Google, University of Illinois Urbana-Champaign · 2610.00964 · ▲ 32
- Multimodal Flow: Unified Flow Modeling of Language and Vision in Embedding Spaces Huazhong University of Science and Technology, Beijing Jiaotong University, Horizon Robotics · 2609.40362 · ▲ 30
- Argo-Bench: Evaluating Data Agents on Enterprise-Scale Workflows TextQL · 2610.02122 · ▲ 30
- Ego2Act: Evaluating Goal-Directed Manipulation in Egocentric Video Generation Mohamed bin Zayed University of Artificial Intelligence, Independent researchers, Nanyang Technological University · 2610.01092 · ▲ 27
- PhysVista: Benchmarking Physical Intelligence in VLMs via a Perception-Reasoning-Assessment Loop University of Science and Technology of China, ByteDance, City University of Hong Kong · 2610.00559 · ▲ 24
- Smaller Models, Better Rejects: Preference Distillation Scaling LinkedIn, University of California, Davis, Arizona State University · 2609.38987 · ▲ 21
- Omni-Embed-Mini: Binding Modalities Without Forgetting via Dense Distillation Mohamed bin Zayed University of Artificial Intelligence · 2610.02148 · ▲ 20
- PixelDense: Dense Prediction as Representation Alignment for Pixel Diffusion University of Virginia, Michigan State University, Adobe · 2610.00483 · ▲ 19
- When Users Change Their Minds: Measuring and Repairing Intent Drift in LLM Agents Hong Kong University of Science and Technology, Chinese University of Hong Kong, Lightspeed · 2609.32520 · ▲ 17
- Align Then Reason: A Multimodal Lip-Sync Judge for Dubbing University of Maryland, Netflix · 2610.00825 · ▲ 17
- Where-OPD: Spatially Guided On-Policy Self-Distillation of MLLMs with Synthetic Scenes Valeo AI, Sorbonne University, Institut universitaire de France · 2610.02117 · ▲ 17
- Pretrain Once, Route Anywhere: Towards a Foundation Model for LLM Routing Nanjing University · 2609.37362 · ▲ 16
- Benchmarking and Enhancing Skill-Level Memory for Partially Observable Robotic Manipulation University of Science and Technology of China, Shanghai AI Laboratory, Zhejiang University · 2609.38886 · ▲ 16
- Better Supervision Is Nearby: Neighborhood On-Policy Self-Distillation Chinese University of Hong Kong, Shenzhen, Meituan, Peking University · 2609.39687 · ▲ 16
- DataMagic: Authoring Data Videos through Declarative Multi-Agent Orchestration Hong Kong University of Science and Technology, Guangzhou · 2609.33403 · ▲ 15
- OmniSeek: Native Tool Integration for Multi-turn Audio-Visual Reasoning University of California, Davis, Adobe · 2610.02181 · ▲ 15
- OpenTumorBoard: A Real-World Benchmark of Multidisciplinary Tumor Board Discussion Trajectories Rice University, University of Illinois Urbana-Champaign, University of Washington · 2609.32810 · ▲ 14
- AgSpec: Pushing the Limits of Retrieval-Based Speculative Decoding in Coding Agent Pipelines KAIST · 2610.01108 · ▲ 13
- RLE-Bench: A Qualifying Exam for Coding Agents as Robot Learning Engineers Harvard University, Georgia Institute of Technology · 2609.34210 · ▲ 12
- Learning What to Recall: Adaptive Multi-Cue Episodic Memory for World Models KAIST, Sony, Imperial College London · 2609.34677 · ▲ 12
- Removing the NEEDLE in the Haystack: Backdoor Removal in LLMs via Weight Orthogonalisation Locai Labs, University College London · 2610.00348 · ▲ 12
- ScholarCatalyst: A Benchmark for Retrieving Papers That Inspire New Research Stanford University, Seoul National University, University of Washington · 2610.02202 · ▲ 12
- Do Audio LLMs Listen Before They Act? Diagnosing Acoustic-Context Gating in Voice Agents Hong Kong University of Science and Technology, Lightspeed · 2609.32536 · ▲ 11
- Devils in Question Relay: Source-Conditioned Relay Steering to Mitigate Hallucinations in Audio-visual Large Language Models Harbin Institute of Technology, Shenzhen, Peng Cheng Laboratory, Fudan University · 2609.37568 · ▲ 11
- LOCI: Spatial Linear Memory for Streaming World Models Mohamed bin Zayed University of Artificial Intelligence, Pinscreen · 2609.40222 · ▲ 11
- Memorizon: Training World Models Beyond Their Context Window Mohamed bin Zayed University of Artificial Intelligence · 2610.00544 · ▲ 11
- SemanTok: Predictable Semantic Tokens for Efficient Autoregressive Video Generation Stability AI, Karlsruhe Institute of Technology · 2610.00686 · ▲ 11
- Generalization Is Stability, Not Accuracy: Multi-Axis Evaluation of LLMs Technion, Israel Institute of Technology · 2610.01428 · ▲ 11
- VTR-Bench: A Systematic Benchmark for Evaluating Visual Text Rendering in Video Generation City University of Hong Kong, Hong Kong University of Science and Technology, Guangzhou, Westlake University · 2610.01499 · ▲ 11
- Rules to Tools: Executable Checks for LLM Agents in Scientific Computing Carnegie Mellon University, DP Technology · 2610.00313 · ▲ 10
- Explore Broadly, Reason Sharply: Push Small Models toward the Frontier via Sampling Georgia Institute of Technology, University of Texas at El Paso, Morgan Stanley · 2609.38104 · ▲ 9
- Prompt2Skill: Unsupervised Skill Optimization From Natural Language Instructions Vanderbilt University, University of Southern California, Adobe · 2609.38593 · ▲ 9
- Does Native 3D Texture Generation Necessarily Require 3D Assets for Training? Chinese University of Hong Kong, Tencent, Tsinghua University · 2609.34621 · ▲ 8
- FlexRouter: Learning Complementary Model Sets for Flexible LLM Routing Virginia Tech, Independent researchers, University of Southern California · 2609.38585 · ▲ 8
- Latent-Foresight: End-to-End Learning Predictable Representations for Latent World Models Archimedes at the Athena Research Center, Valeo AI, National Technical University of Athens · 2610.01942 · ▲ 8
- KaliBench: A Fine-Grained Benchmark for Cybersecurity Tool Use on Kali Linux with Runtime-Free Verifiable Rewards Khalifa University, University of Western Australia · 2610.02206 · ▲ 8
- Joint and Cross-Modal Video-Audio Generation and Editing: A Unified Formulation and Design Taxonomy University of Massachusetts Amherst, University of Texas at Dallas, Virginia Tech · 2609.34381 · ▲ 7
- MemFold: Learning Compact Soft Memory for Long-Context Personalization via On-Policy Optimization University of California, Santa Barbara, University of North Carolina at Chapel Hill, Harvard University · 2609.36435 · ▲ 7
- Personalized Image Generation with Reasoning and Reflection Vanderbilt University, Adobe, University of Maryland · 2610.00737 · ▲ 7
- It Takes Workflows to Evolve Better Workflows William and Mary, NEC Corporation of America, University of Illinois Urbana-Champaign · 2610.01026 · ▲ 7
- When Does Correction Become Repair? Mechanistic Auditing of Internal Interventions in Tool-Using LLMs University of Chinese Academy of Sciences, University of Birmingham · 2609.36138 · ▲ 6
- Controlled Decoding Attacks on Black-Box LLMs University of Southern California, Adobe, Amazon · 2609.36956 · ▲ 6
- DexPolicy: Scheduled Exploration for Trajectory-Guided Dexterous Manipulation Peking University, AI2 Robotics · 2610.00360 · ▲ 6
- JevSpawn: Adaptive Agentic Inference through Compositional Action Spaces Fudan University, Shanghai Jiao Tong University, Shanghai Innovation Institute · 2610.00437 · ▲ 6
- Replacing Large Language Models with Jev Decision Models for Low-Latency Edge Service Orchestration University of Technology Sydney · 2609.22753 · ▲ 5
- OTRetarget: Joint Robot and Object Motion Retargeting via Optimal Transport Inria, École Normale Supérieure, Stanford University · 2609.36602 · ▲ 5
- Honeycomb: Constant-Size Scene Memory Representation for Video World Models Harvard University, National University of Singapore, Massachusetts Institute of Technology · 2609.37690 · ▲ 5
- Pay for the Fault, Not the Flow: Label-Free In-Flow Multi-Agent Workflow Optimization William and Mary, NEC Corporation of America · 2610.01017 · ▲ 5
- Before It Fades: Reinforcing Temporal Representations at Inference Time in VideoLLMs KAIST · 2610.01595 · ▲ 5
- Keyword Harnesses Fail Open: A Cheap Diagnostic Ladder for Tool-Use Claims in Small Language Models Independent researchers · 2610.02142 · ▲ 5
Thursday, October 1, 2026 · 86
- LoopVL: Recurrent Visual Intelligence Renmin University of China, Baidu, Shanghai Jiao Tong University · 2609.38426 · ▲ 454
- The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation University of Science and Technology of China, Rutgers University, Zhejiang University · 2609.36484 · ▲ 421
- False Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents Rutgers University, Independent researchers, University of California San Diego · 2609.39102 · ▲ 393
- UniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvement Stanford University, Johns Hopkins University, Independent researchers · 2609.38721 · ▲ 277
- Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual Evidence University of Alabama at Birmingham, Amazon · 2609.34563 · ▲ 209
- AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks Beijing Academy of Artificial Intelligence · 2609.38288 · ▲ 128
- Mid-Harness: Scaling Actions Between Model and Harness for Terminal Agents NVIDIA, KAIST · 2609.39982 · ▲ 112
- EvoDuet: Bilevel Co-Evolution of Web Searching and Task Solving for Scientific Discovery University of Minnesota, KAIST, Seoul National University · 2609.40340 · ▲ 106
- WorldAuditBench: Interactive 3D World Auditing with Multimodal Agents UC Santa Barbara, Massachusetts Institute of Technology, IBM · 2609.40325 · ▲ 95
- RSIGame: Autonomous Agentic Game Development with Recursive Self-improvement University of California San Diego, ByteDance, Carnegie Mellon University · 2609.39045 · ▲ 83
- Learning Meta-Skills for Agent Harness Design in Test-Time AI4AI Apodex, University of Illinois Urbana-Champaign · 2609.38143 · ▲ 80
- EVOKE: Eliciting World Knowledge in Agents for Transferable Decision-Making Shanghai Jiao Tong University, Hong Kong Polytechnic University, Eastern Institute of Technology, Ningbo · 2609.38334 · ▲ 70
- Thinking Outside the Box: Can Language Models Rely on External Guidance Selectively? Southern University of Science and Technology, City University of Hong Kong · 2609.39578 · ▲ 61
- Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering KAIST, ETH Zurich, Google · 2609.38177 · ▲ 60
- OSWorld-Science: A Benchmark of Computer Use Agents for Learning and Using Scientific Software Tsinghua University, University of California, Los Angeles, RIKEN AIP · 2609.39903 · ▲ 56
- More Choices, Fewer Decisions: Ordinal-Scale Bias in JEV-like Direct-Decision Models Jilin University · 2609.38827 · ▲ 54
- AIM: Agentic Idea Management for Automated Research Google, University of Wisconsin-Madison · 2609.38445 · ▲ 50
- Agent Error Dataset: Scaling 50,000 Error--Diagnosis Pairs for Failure Analysis and Error-Aware Post-Training Apodex · 2609.40111 · ▲ 47
- LANTERN: Illuminating Hidden Mathematical Knowledge in Language Models arXiv · 2609.32264 · ▲ 46
- Unmask the State: When Does State Adaptation Matter for Masked Diffusion Language Models Seoul National University · 2609.33355 · ▲ 42
- DC-SAE: Deep Compression Semantic Autoencoder for Faster Diffusion Convergence Peking University, Singapore University of Technology and Design · 2609.39222 · ▲ 37
- It's Not What the Image Shows: Irrelevant Context Destabilises VLM Judges Without Informing Them Technion · 2609.37863 · ▲ 36
- DyRAD: Radar Novel View Synthesis for Dynamic Driving Scenes Technion, Cornell Tech, NVIDIA · 2609.39841 · ▲ 36
- Breaking Babel: A Self-Evolving Multi-Agent System for Long-Form Subtitle Translation University of Illinois Urbana-Champaign, Amazon · 2609.38660 · ▲ 35
- TERRA: Terrain-Aware Reconstruction, Retargeting and Control for Musculoskeletal Locomotion EPFL · 2609.38653 · ▲ 33
- Systematically Exploring the Capabilities of GPT-6 Astra as Embodied Policies Galbot · 2609.38537 · ▲ 30
- ThinkV2V: Unleashing the Reasoning Capability of MLLMs for Instruction-Guided Video Editing Chinese University of Hong Kong, Zhejiang University, ByteDance · 2609.38541 · ▲ 30
- BiasReducer: Adaptive Bias Mitigation for Reward Models Carnegie Mellon University, Chinese University of Hong Kong, Shenzhen, New Jersey Institute of Technology · 2609.32720 · ▲ 24
- PivotOPD: Learning to Recover from Pivotal Mistakes in Multi-Turn Agents Princeton University, NVIDIA, University of Maryland · 2609.40285 · ▲ 21
- Scaling Laws for Looped Mixture of Experts Meta · 2609.40316 · ▲ 21
- PixelUMM: Encoder-Free Unified Image and Video Understanding and Generation NVIDIA, University of Waterloo · 2609.38597 · ▲ 20
- CUA-SWE: When Computer-Use Agents Meet Visual Software Engineering Carnegie Mellon University, University of Southern California, University of Wisconsin-Madison · 2609.32600 · ▲ 17
- The Low-Rank Structure of VLA Reinforcement Learning Seoul National University · 2609.34599 · ▲ 16
- MILO: Automated Harness Discovery via Orchestrated Multi-Agent Evolution Georgia Institute of Technology, Amazon, Carnegie Mellon University · 2609.38349 · ▲ 16
- A2Z GameSpec-Bench: How Faithfully Can Coding Agents Generate Games from Game Design Specifications? Krafton, KAIST, Korea National University of Arts · 2609.39564 · ▲ 14
- Physis-Lang: Self-Evolving Language as a Physical Representation for Video World Model NVIDIA, Massachusetts Institute of Technology, University of Oxford · 2609.40358 · ▲ 14
- DuoOPD: Learning from Joint Teacher-Student Outcomes for Multi-Task On-Policy Distillation University of Science and Technology of China, Hong Kong Polytechnic University, University of Hong Kong · 2609.33711 · ▲ 13
- Rubric Rewards from Item Response Theory University of British Columbia, Microsoft · 2609.35646 · ▲ 13
- RoboCoach: World Models as Active Coaches for Compositional Robot Skills Renmin University of China, Tsinghua University, University of Nottingham · 2609.39685 · ▲ 13
- I Have a Stream: Making Self-Supervised Learning Work on Continuous Video University of Zagreb, University of Technology Nuremberg · 2609.40333 · ▲ 12
- Tacit-TTS: From Autoregressive Decoding to Masked Prediction for Efficient Transcript-Free Voice Cloning Dolby Laboratories · 2609.38658 · ▲ 11
- Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior University of Sheffield, Mohamed bin Zayed University of Artificial Intelligence, Hugging Face · 2609.39827 · ▲ 11
- SeLMRoute: Probabilistic Semantic Evidence for Large Language Model Routing Indigma Innovations, Democritus University of Thrace, Athena Research Center · 2609.34736 · ▲ 10
- AdviSD: Learning to Advise Frontier LLMs via Targeted Multi-Turn Self-Distillation Google, University of Southern California · 2609.38142 · ▲ 10
- The Evolution of Attention in Large Language Models: Mechanisms, Trade-offs, and Emerging Trends Alibaba · 2609.39661 · ▲ 10
- WUSH-KV: KV Cache Quantization with Data-Adaptive Transforms Institute of Science and Technology Austria, Red Hat AI, ETH Zurich · 2609.38121 · ▲ 9
- PatchHolmes: Agentic Patch Retrieval via Listwise Selection Stevens Institute of Technology · 2609.38807 · ▲ 9
- CoEvoWhen: Policy-Tool Coevolution for Ultra-Long Video Temporal Grounding Zhejiang University · 2609.40048 · ▲ 8
- SkillGym: Training Skill-Use Agents with Automatic Verifiable Environment Generation Mohamed bin Zayed University of Artificial Intelligence · 2609.37539 · ▲ 7
- Overcoming Scaling Limits in On-Policy Self-Distillation for LLM Reasoning North South University, KAIST, Andria Labs · 2609.37915 · ▲ 7
- Safety of Latent Communication in Multi-Agent Systems CISPA Helmholtz Center for Information Security · 2609.39788 · ▲ 7
- SkillSeek: Revisiting Agent Skill Retrieval at Marketplace Scale Stevens Institute of Technology, Independent researchers · 2609.38822 · ▲ 6
- Not Every Token Is Worth Distilling: Selective Supervision for Direct-OPD East China Normal University · 2609.29142 · ▲ 5
- NavHarness: Towards Lifelong Embodied Navigation Adelaide University, Australian Institute for Machine Learning, CSIRO Data61 · 2609.34276 · ▲ 5
- Decompose Radicals, Then Reward: Fine-Grained Inspection for Accurate Chinese Text Rendering Fudan University, Shanghai Key Laboratory of Multimodal Embodied AI · 2609.37569 · ▲ 5
- Training LLM Judges from Language Feedback via Position-Selective Self-Distillation Georgia Institute of Technology, Amazon, University of California San Diego · 2609.38792 · ▲ 5
- Beyond the Remembered World: Predictive 4D Belief for Persistent Navigation in Evolving Worlds Zhejiang University, University of California San Diego, Deeprobotics · 2609.39166 · ▲ 5
- Game-Guided Skill Discovery through Self-Play for Playable Agent Control Georgia Institute of Technology, Simon Fraser University, NVIDIA · 2609.40137 · ▲ 5
- Removing Timing Shortcuts Improves Non-Invasive Brain-to-Text University of Oxford · 2609.40359 · ▲ 5
- Understanding Multimodality in Generative Behavioral Cloning arXiv · 2605.22493 · ▲ 4
- See it, Say it, Sorted: Mechanistic Diagnosis and Parameter-Space Mitigation of Emergent Misalignment in LLMs East China Normal University, University of Birmingham, Alibaba · 2609.34970 · ▲ 4
- CheatBench: Measuring Reward Gaming in AI Agents Center for AI Safety · 2609.36308 · ▲ 4
- Mitigating the Length-Scaling Tax with Online Distillation ByteDance, Tongji University, Peking University · 2609.38854 · ▲ 4
- DAGent: Evaluate-then-Grow Planning for Deep Research Agents New York University, New York University Shanghai · 2609.39154 · ▲ 4
- How Much Is an AI Token Worth? Scaling Laws for Wild AI-Generated Web Text University of Maryland, Pangram Labs · 2609.40295 · ▲ 4
- Label-free steering: Compressing test-time reinforcement learning into bias-only subspaces University of Turku, ELLIS Institute Finland · 2609.18587 · ▲ 3
- The Endless Exam: Mathematical Constructions from Today's Models toward Superintelligence University of Texas at Arlington · 2609.24555 · ▲ 3
- BIABench: Evaluating AI agents on real-world bioimage analysis tasks University of Notre Dame, Leibniz Institute for Analytical Sciences, University of Tartu · 2609.34274 · ▲ 3
- Dating the Model: Hidden Dates in System Prompts Affect LLM Evaluation Johannes Gutenberg University Mainz, Universidad Iberoamericana, University of Colorado Boulder · 2609.36931 · ▲ 3
- Evaluating Bounded Autonomy in Regulated Agentic AI: A Diagnostic Harness with Constitutional Rewards, Escalation Labels, and Runtime Governance Skelf Research · 2609.37501 · ▲ 3
- SpatialCORE: Confidence-Aware Grounded Spatial Reasoning in Large Vision--Language Models Wayne State University, Henry Ford Health, Ohio State University · 2609.38716 · ▲ 3
- Working Around the Compute Ceiling: Byte-Exact Memory in Galahad Makes LLM Reading a One-Time Cost LLM Reading a One-Time Cost Corbenic AI · 2609.39358 · ▲ 3
- EviRover: Reinforcing Agentic Perception Beyond a Glance Chinese University of Hong Kong, Fudan University · 2609.40230 · ▲ 3
- Decision-Oriented Recommendation Reranking: An Empirical Study of Jev Singapore Management University, Meta · 2609.40241 · ▲ 3
- Aligning One-Step Generative Models with Reward-Weighted Transport Distillation ByteDance, University of California, Berkeley · 2609.30840 · ▲ 2
- SlideDP: Scaling Host-Resident LLM Fine-Tuning Across Multiple GPUs Hong Kong University of Science and Technology, Guangzhou, Beijing Academy of Artificial Intelligence · 2609.34162 · ▲ 2
- Almost Human, Except When It Matters: VoxParity and the Decisions a Voice Should Change Independent researchers · 2609.35922 · ▲ 2
- Prompted Identity Degrades Cooperation in Multi-Agent LLM Systems Sapienza University of Rome · 2609.35928 · ▲ 2
- ATLAS: Aligned Transport of Latent Structure for Reliable World Model Planning Pennsylvania State University · 2609.36333 · ▲ 2
- The Geometry of Inference in Transformer Residual Streams University of Luxembourg, London Institute for Mathematical Sciences · 2609.37824 · ▲ 2
- Retrieval Capacity of Self-Attention Under Competition University of Luxembourg, London Institute for Mathematical Sciences · 2609.37879 · ▲ 2
- Soft Spatial Reasoning Wayne State University, Henry Ford Health, Ohio State University · 2609.38717 · ▲ 2
- Learning Chaos Without Seeing Chaos: Extrapolation of Global Dynamics in Autoregressive Transformers Ludwig Maximilian University of Munich, Technical University of Munich, Munich Center for Machine Learning · 2609.38814 · ▲ 2
- Who Said What, and Will It Be Remembered? Evaluating Persistent Speaker Attribution Across Meetings Foyer · 2609.39344 · ▲ 2
- LEAP: Learned Block-wise Evidence Retrieval for Long Audio-Video Perception Northeastern University, Futurewei Technologies · 2609.39938 · ▲ 2
- MemLife: Curating and Reasoning over Long-Term Egocentric Video Memories Meta, University of Virginia, University of Illinois Urbana-Champaign · 2609.40195 · ▲ 2
Wednesday, September 30, 2026 · 92
- Raven: The Harness of Harnesses for Composable Agentic Intelligence EverMind AI · 2609.33439 · ▲ 550 · Deep dive
- MaLiang-Harness: A Programmable Path to Image and Video Generation National University of Singapore, Fudan University, Tencent · 2609.34309 · ▲ 405
- In-Context Learning for Robots: Methods and Applications Knowin AI, Hong Kong University of Science and Technology, Guangzhou, Chinese University of Hong Kong · 2609.36012 · ▲ 379 · Deep dive
- Scaling Properties of Same-Family On-Policy Distillation Zhejiang University, Kuaishou · 2609.32722 · ▲ 315
- Omni-IO Skills: Harnessing Your Agent Omni-Native National University of Singapore, University of Oxford · 2609.31847 · ▲ 287 · Deep dive
- PanoVLN: Towards Effective Panoramic Vision-and-Language Navigation Zhejiang University, University of Hong Kong · 2609.34759 · ▲ 160
- VoxMem: Benchmarking Multimodal Memory in Large Audio Language Models University of Melbourne, UNSW Sydney · 2609.32607 · ▲ 149
- LEGO-Anything: Coding Agents for 3D Scene Reconstruction University of Maryland, Amazon · 2609.36380 · ▲ 129
- What Makes World Action Models Generalize? An Empirical Study of Test-Time Future Modeling Tsinghua University, University of Science and Technology of China, Beijing Institute of Technology · 2609.34981 · ▲ 127
- Beyond the Timeline: Augmenting Long-Video Memory with Grounded Entity Biographies University of Illinois Urbana-Champaign, Amazon · 2609.38155 · ▲ 105
- Think Before You Score: Thinking Reward Model for Visual Generation Hangzhou Dianzi University, Institute of Automation, Chinese Academy of Sciences, Peking University · 2609.37372 · ▲ 98
- Periodic Weak Spots: Phase Sensitivity from Chunked KV-Cache Compression ByteDance, Princeton University, Stanford University · 2609.36322 · ▲ 96
- Follow the Entities: A Corpus Map for Agentic Search KAIST, Microsoft · 2609.37226 · ▲ 95
- SAKI: Maximal-Coupling-Routed Teacher Supervision for On-Policy Distillation Meituan, KTH Royal Institute of Technology · 2609.36601 · ▲ 91
- LLMs are General Asynchronous Agents Yandex, Together AI, HSE University · 2609.35427 · ▲ 77
- EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments? Tsinghua University, Zhiman, Chongqing University · 2609.37686 · ▲ 68
- Beyond Dyadic Memory: Interaction-Aware Multimodal Memory with Adaptive Agentic Retrieval for Multi-Party Spoken Conversations Zhejiang University, Meituan · 2609.32522 · ▲ 63
- LongCat-DeepResearch Technical Report Meituan · 2609.36071 · ▲ 63
- Learning Beyond What You Sample: Off-Policy-Aware Cross-Model Trajectory Exchange for RLVR KAIST, AITRICS · 2609.37868 · ▲ 59
- OmniTaskonomy: When Does Visual Generation Improve Visual Understanding? University of California, Berkeley, Duke University, Carnegie Mellon University · 2609.38079 · ▲ 52
- ROSS: Relearning from Self-Generated Rollouts through Selective Supervision arXiv · 2609.35954 · ▲ 46
- WorldAttention: An Efficient Attention Architecture for Interactive Video World Models Alibaba, Zhejiang University, Hong Kong University of Science and Technology · 2609.34606 · ▲ 44
- APM-Bench: Benchmarking Cross-session Persistent Memory for Egocentric Streaming Video Assistants Shanghai Jiao Tong University, Eastern Institute of Technology, Ningbo, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences · 2609.37559 · ▲ 43
- HybridCUA: Learning to Orchestrate GUI and CLI for Computer-Use Agents Zhejiang University, Peking University, Tsinghua University · 2609.38008 · ▲ 43
- EVO-WAM: Evolving World Action Models through Video-Action Verification Harbin Institute of Technology, Shenzhen, SLAI, Tsinghua University · 2609.38057 · ▲ 40
- LongLive-Plug: Once-for-All Distillation for Video Generation NVIDIA · 2609.38154 · ▲ 40
- SoL-Refiner: Speed-of-Light One-Step Refinement for High-Resolution Video NVIDIA · 2609.37969 · ▲ 39
- Context Language Models University of Washington, Meta, Massachusetts Institute of Technology · 2609.37725 · ▲ 38
- Marathoner: Ultra-Long-Horizon Autonomous Intelligence University of Macau, Peking University, Ant Group · 2609.34378 · ▲ 37
- Asking for What Was Never Requested: Horizontal and Vertical Proactivity in Agents IBM, Weizmann Institute of Science · 2609.37236 · ▲ 37
- EasyPPO: Stabilizing the Critic Is Key University of California, Berkeley, Princeton University · 2609.36802 · ▲ 35
- ANTMAN: Adaptive Need Tracking for Multi-Agent Navigation in Large Information Spaces University of Illinois Urbana-Champaign · 2609.33326 · ▲ 32
- Anisotropic Representations Improve Planning in JEPA World Models Seoul National University, Ewha Womans University · 2609.37441 · ▲ 32
- FocusVTC: Efficient and High-Performance Visual Text Compression with Adaptive Resolution Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Shanghai Jiao Tong University · 2609.36651 · ▲ 29
- Omni-Decision: Evidence-Ledger Planning for Omni-Modal Agents Institute of Neuroscience, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Alibaba · 2607.11433 · ▲ 28
- WorldLine: Action-Driven Visual Simulation for Robotic Manipulation Hong Kong University of Science and Technology, Joy Future Academy, Peking University · 2609.38059 · ▲ 27
- CrossBFM: Distilling a Shared Latent Behavior Space Across Humanoid Embodiments VinRobotics, National University of Singapore, Technical University of Darmstadt · 2609.38087 · ▲ 25
- Reasoning with Image Generation Technical University of Darmstadt, hessian AI · 2609.16409 · ▲ 22
- When Does Dense Retrieval Need Asymmetric Geometry? A Bias-Variance Theory of Shared and Dual Projections Hong Kong Polytechnic University · 2609.32488 · ▲ 22
- Chinese-Jev: Bringing System One Model to Chinese-Language Tasks Fudan University, National University of Singapore, University of Chinese Academy of Sciences · 2609.36965 · ▲ 22
- HiRAE: Hierarchical Representation Autoencoding with Residual Budgets Peking University, Agibot Research, Tsinghua University · 2609.37775 · ▲ 22
- On the Off-Policy Teacher in On-Policy Distillation Washington University in St. Louis, Amazon, Carnegie Mellon University · 2609.38360 · ▲ 21
- EmoRES-TTS: Residual-Enhanced Vector Steering for Emotional Speech Generation Reality Labs at Meta, Meta, National Taiwan University · 2609.38157 · ▲ 20
- Adversarial Training for Pixel Diffusion University of California San Diego, Adobe, University of California, Merced · 2609.38170 · ▲ 20
- TabFM: A Zero-Shot Foundation Model for Tabular Data Google · 2609.37959 · ▲ 19
- AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation University of Tokyo · 2609.35530 · ▲ 17
- Scheduling Recursive Reasoning in Looped Transformers Southern University of Science and Technology, Chinese University of Hong Kong, Shenzhen, Shenzhen Loop Area Institute · 2609.36653 · ▲ 17
- Selecting The Most Informative Tokens in Natural Language Autoencoders University of Turin, University of Southern Denmark · 2609.37040 · ▲ 16
- Routing Should Pay for Itself: Sparse Supervision for Economical LLM Routing Nanjing University, Hong Kong University of Science and Technology, SinapisAI · 2609.37402 · ▲ 16
- Language Models Are "Insecure" Reporters Massachusetts Institute of Technology, Google, Harvard University · 2609.36139 · ▲ 14
- Can Agents Design Libraries for Agents? University of Wisconsin-Madison, Massachusetts Institute of Technology, Snorkel AI · 2609.36730 · ▲ 11
- TabFM-Auto: Self-Evolving Pipelines for Tabular Foundation Models Google, University of Southern California, Harvard University · 2609.37989 · ▲ 11
- VideoLoop: Looped Working Memory Against Semantic Thrashing in Long-Form Video Agents University of Rochester, Microsoft · 2609.38119 · ▲ 11
- StoryEngine: A State-Grounded Agentic Framework for Video Storytelling Tencent, National University of Singapore · 2609.33627 · ▲ 10
- Fractional State Space Transition for Long Sequence Modeling Huawei · 2609.36314 · ▲ 10
- What Makes Recurrence Effective in Looped Language Models? Chinese University of Hong Kong, Mohamed bin Zayed University of Artificial Intelligence · 2609.36636 · ▲ 10
- Where the Model Changes Its Mind: Hindsight-Divergence Localization for Efficient Reinforcement Learning with Verifiable Rewards NVIDIA, University of Wisconsin-Madison, University of California, Berkeley · 2609.36864 · ▲ 9
- TGRL: Temperature-Grouped Reinforcement Learning for Efficient Exploration in LLMs University of Chinese Academy of Sciences, Meituan, Institute of Automation, Chinese Academy of Sciences · 2609.33589 · ▲ 8
- FurE: Efficient Instance-Specific 3D Fur Reconstruction without Animal-Fur Datasets Johns Hopkins University · 2609.35770 · ▲ 8
- EpiCon: Collective Agent Learning through Co-Evolving Multimodal Memory IBM, University of Rochester, Massachusetts Institute of Technology · 2609.37923 · ▲ 8
- ALICE: In-context, Zero-shot, Mutual Information Estimation EURECOM · 2609.34962 · ▲ 7
- Beyond Selection: Token Parameterization for Extreme Visual Token Compression Zhejiang University · 2609.35232 · ▲ 7
- Jev thinks "I don't know'', but doesn't say it: Introducing Sys1Cal-v1 Dataset for Probability Calibration little g.ai, Sant'Anna School of Advanced Studies, University of Pisa · 2609.35342 · ▲ 7
- Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt Injection Peking University, National University of Singapore, BYD · 2609.35932 · ▲ 7
- Learning Dynamics of Continual Learning: A Unified View of Data Attribution, Forgetting, and Plasticity Loss OATML, University of Oxford, University of British Columbia · 2609.33620 · ▲ 6 · Deep dive
- AutoDataBench: Can Agents Write the Data That Feeds the Self-Improvement Loop? Hong Kong University of Science and Technology, SLAI, NTU · 2609.35025 · ▲ 6
- Memory Is a Derivation: The Distributed-Evidence Paradox in Long-Term Agents New York University · 2609.36130 · ▲ 6
- PreviewDiff: Multimodal Critic-Guided Search over Diffusion Latents Massachusetts Institute of Technology, Google, University of California, San Francisco · 2609.36199 · ▲ 6
- Act First, Reason Later: Accelerating On-Policy Distillation for Multi-Turn Agents via Reference-Conditioned Inverse Dynamics Southern University of Science and Technology, Zhongguancun Academy, Hong Kong Polytechnic University · 2609.36608 · ▲ 6
- Real2Gym: Building Gyms from Videos, Bringing Skills to Robots Shanghai AI Laboratory, Shanghai Jiao Tong University, Zhejiang University · 2609.37089 · ▲ 6
- WISE-ATTA: When to Ask for Labels in Budgeted Active Test-Time Adaptation CISPA Helmholtz Center for Information Security · 2609.37687 · ▲ 6
- Breaking the Uniformity Trap: Scaling Video Diffusion Model via SplitMoE University of Chinese Academy of Sciences, ByteDance, Canva Research · 2609.38140 · ▲ 6
- LIFT: Layout-In-Future Video Generation under Large Viewpoint Change via On-Policy Self-Distillation University of California San Diego, University of Virginia, Meta · 2609.38146 · ▲ 6
- Scaffolding Minds: Optimizing Latent Visual Target Representations for Multimodal Reasoning Google, University of California San Diego · 2608.19669 · ▲ 5
- CaptchaArena: A Large-Scale, Fine-Grained Dataset for Training Computer-Use Agents on Interactive CAPTCHAs Columbia University, Zhejiang University, University of Rochester · 2609.31957 · ▲ 5
- Pretraining Transformers with Quantized Softmax in Attention University of Illinois Urbana-Champaign · 2609.33591 · ▲ 5
- Org-Agent: Beyond Personal Assistants Towards Organizational Agents Hong Kong Polytechnic University · 2609.34392 · ▲ 5
- Preference-Guided Adaptation for Open-Vocabulary Semantic Segmentation via Prompt Disagreement KAIST · 2609.34528 · ▲ 5
- PMOPD: Task Ordering, Cycling, and Parameter-Update Subspace Protection in Multi-Teacher On-Policy Distillation Ant Group · 2609.34605 · ▲ 5
- Understanding On-Policy Distillation: A Mechanistic Interpretability Perspective via Sparse Crosscoders University of Hong Kong, University of Cambridge, Shenzhen Loop Area Institute · 2609.35210 · ▲ 5
- Principled Thoughts for Latent Recursive LLM Systems University of British Columbia · 2609.36159 · ▲ 5
- StructRL: Online Structured Reinforcement Learning for Long-Horizon Vision-Language-Action Tasks Pennsylvania State University, Amazon · 2609.36352 · ▲ 5
- Improved Distributional Diffusion Models LMU Munich, Munich Center for Machine Learning, Google · 2609.37147 · ▲ 5
- PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers Nanyang Technological University, Nanjing University of Posts and Telecommunications, Nanjing University · 2609.32429 · ▲ 4
- AnyStep-WAM: Budget-Aligned Distillation and Adaptive Inference for World Action Models arXiv · 2609.33748 · ▲ 4
- PlaylistEval: Can Video-Language Judges Be Trusted at Day Scale and Beyond? KAIST · 2609.34314 · ▲ 4
- One Proposal for Every Margin: Zero-Shot Amortized Sequential Importance Sampling for Binary Matrices Tsinghua University · 2609.35514 · ▲ 4
- Hyperspherical Semantic Trajectory Analysis: Mapping Technological Diffusion across Academic Preprints, Patent Signals, and Compute Scaling Asia School of Business · 2609.35845 · ▲ 4
- Persistence Forcing: Exploiting Feature Specialization in Pixel-Space Diffusion Nanyang Technological University, KTH Royal Institute of Technology, Hebei University of Technology · 2609.36014 · ▲ 4
- SEAD: A State-Based Perspective on Attack and Defense in Tool-Using Agents Georgia Institute of Technology · 2609.34518 · ▲ 3
- AgentTell: Behavioural Side-Channel Leakage in Browser-Use Agents BRAC University, Qatar Computing Research Institute · 2609.32915 · ▲ 2
- Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation Amazon, University of California, Berkeley, Carnegie Mellon University · 2609.38172
Tuesday, September 29, 2026 · 98
- Post-Training Leaves Behavioral Shadows on Unrelated Decisions Peking University, Georgia Institute of Technology, ShanghaiTech University · 2609.29233 · ▲ 271 · Deep dive
- YuE2: Unifying Symbolic and Audio Music Generation at Frontier Quality Hong Kong University of Science and Technology, HKGAI, M-A-P · 2609.33757 · ▲ 241 · Deep dive
- VisionHOPE: Visual Backbones as Self-Modifying Learning Systems Mininglamp Technology, Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences · 2609.33325 · ▲ 215
- Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation Nanyang Technological University, Yale University, University of Manchester · 2609.35347 · ▲ 176
- Groupwise Agentic Grading and Advantage Redistribution for Code Agent RL Xiaomi, Renmin University of China, Peking University · 2609.32577 · ▲ 126 · Deep dive
- Self-Evolving Coding Agents: From Digital Programs to Physical-World Intelligence HexaFuture · 2609.35432 · ▲ 106
- Duplex-MPE: Benchmarking Multi-Party Interaction in Full-Duplex Dialogue Peking University, Renmin University of China, Tsinghua University · 2609.31948 · ▲ 89
- MassAlloc Attention: Let Attention Allocate Its Own Compute Hong Kong University of Science and Technology, Guangzhou, Beijing Academy of Artificial Intelligence, Université Paris Cité · 2609.32712 · ▲ 74
- TraceDance: An Automated System for Building Agent Behavior Benchmarks from Real-World Agent Deployment Traces ByteDance, University of Illinois Chicago · 2609.33295 · ▲ 69
- CoWindow Attention: Full Causal Coverage Is a Collective Property Hong Kong University of Science and Technology, Guangzhou, Beijing Academy of Artificial Intelligence, Université Paris Cité · 2609.32704 · ▲ 66
- How Far Are We from Removing the Visual Encoder? Scaling Laws for Encoder-Free Multimodal Pretraining Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Tencent · 2609.35457 · ▲ 65
- Improving Test-Time Scaling with Adaptive Looped Transformers Tsinghua University, Yale University · 2609.35748 · ▲ 55
- Learning Native Reflection in Unified Models with Interleaved Reinforcement Learning Nanyang Technological University, Shanghai Jiao Tong University, University of Tokyo · 2609.35767 · ▲ 48
- Surprising Success, Repeated Failure: Entropy-Guided Credit Assignment for Exploration in LLM Reasoning KAIST, DeepAuto AI · 2609.33781 · ▲ 41
- QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents Alibaba, University of Science and Technology of China, Tsinghua University · 2609.33848 · ▲ 41
- Recursive Harness Distillation across Agents for Robot Manipulation Seoul National University, Korea Advanced Institute of Science and Technology · 2609.33378 · ▲ 40
- Knowing When Thinking Is Not Enough: Teaching Small Reasoning Models to Reason Beyond Their Parametric Knowledge KAIST, DeepAuto AI · 2609.34327 · ▲ 40
- EmbodiedMemory-Bench: Benchmarking Embodied Memory for Long-Horizon Embodied Tasks Zhejiang University, Central South University, Institute of Software, Chinese Academy of Sciences · 2609.28236 · ▲ 39
- CompoWorld: Compositional Environment Scaling for General Agents AllSpark Team · 2609.33665 · ▲ 38
- Learning from Teacher Continuations at Student States University of Illinois Urbana-Champaign, Massachusetts Institute of Technology, University of Pennsylvania · 2609.36246 · ▲ 38
- Selecting Diverse SFT Traces Improves Post-RL Generalization University of Illinois Urbana-Champaign, Google · 2609.33780 · ▲ 37
- Skill2Env: Capability-Oriented Environment Synthesis from Skills for General Agents AllSpark Team · 2609.33772 · ▲ 35
- In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion Meta, Nanyang Technological University · 2609.32540 · ▲ 33
- Diffusion Reward Models Tsinghua University, Chinese University of Hong Kong, University of Illinois Urbana-Champaign · 2609.33803 · ▲ 33
- DepthBench: Measuring How Residual Connections Enable More Computational Depth ELLIS Institute Tübingen, Max Planck Institute for Intelligent Systems, Tübingen AI Center · 2609.32534 · ▲ 31
- Learning to Learn from Context: Synthetic Training from Perturbed Public Documents AllSpark Team · 2609.33642 · ▲ 31
- WorldPlay2: Extending Real-Time Interactive World Models in Control and Horizon arXiv · 2609.35560 · ▲ 31
- FlowTool: Controlling Tool Parameter in Image Retouching via Flow Matching Adobe, KAIST · 2609.35673 · ▲ 31
- Rethinking Training-Inference Mismatch in LLM Reinforcement Learning: Where It Arises and How to Correct It Brigham Young University, University of North Carolina at Chapel Hill · 2609.32444 · ▲ 30
- Just MLPs: Efficient Visual State Reconstruction for Multimodal Language Models Nanyang Technological University, Shanghai Jiao Tong University, Fudan University · 2609.34972 · ▲ 30
- RoboFoundry: System-as-Policy Evolution for Self-Learning Embodied Agents Nanyang Technological University, Beihang University, National University of Singapore · 2609.32862 · ▲ 25
- SpatialSpeak: QA-Native Reconstruction with Local and Global Context for Spatial Chain-of-Thought Reasoning Hong Kong University of Science and Technology, Huawei, Harbin Institute of Technology · 2609.33616 · ▲ 25
- An RL View of OPD: Least Square Policy Distillation for Sample-Efficient LLM Reasoning University of North Carolina at Chapel Hill, Brigham Young University, NVIDIA · 2609.35505 · ▲ 22
- Change the Product, Keep the Parameters: Associative Algebra Layers for Transformers DAIMLD, Artificial Intelligence Research Institute · 2609.32814 · ▲ 21
- Hard Vision, Easy Vision: What GPT-6 Astra Reveals Across Computer Vision Mohamed bin Zayed University of Artificial Intelligence, Apertix · 2609.35718 · ▲ 21
- REALM: A Coarse-to-Fine Generative Framework for Embodied Reactive Listening Macquarie University, University of North Texas · 2609.33095 · ▲ 20
- Precise Editing and Flexible Referencing for Interactable Worlds Nanyang Technological University, StepFun · 2609.34470 · ▲ 19
- VideoPhysEdit: Physical Counterfactual Video Editing via Rigid-Body Physical Scene Reconstruction Fudan University · 2609.35134 · ▲ 19
- AdaTutoRank: Learning to Rerank Document Sets via Adaptive Tutoring Optimization for RAG and Deep Research University of Science and Technology of China, Tencent · 2609.32472 · ▲ 18
- TT-VidT: Decoupling the Temporal Axis for Efficient Motion-Centric Video Pretraining National Tsing Hua University, Comfy Org Research, Realiz AI · 2609.33419 · ▲ 18
- Nereus: Adaptive Parallelism for LLM Post-Training Aalto University, Shenzhen University of Advanced Technology, Zhejiang University · 2609.34645 · ▲ 18
- ExpVoyager: Direct Experience Navigation for Dynamic Agent Skill Synthesis Yonsei University · 2609.32630 · ▲ 17
- Structured Residual Connectivity Matters for Diffusion Transformers National University of Singapore, Shanghai Jiao Tong University · 2609.33203 · ▲ 17
- WideSWE: Can Coding Agents Coordinate Changes Across Repositories? Zhejiang University, Tsinghua University · 2609.33382 · ▲ 17
- SciGen-Verifier: A Multimodal Reasoner for Explainable Verification in Scientific Image Generation South China University of Technology, Hong Kong Polytechnic University · 2609.33399 · ▲ 16
- Imprint Reader: From Weight-Update Readout to Behavioral Intervention Shanghai AI Laboratory, Shanghai Jiao Tong University · 2609.35261 · ▲ 16
- SolveEdit: Benchmarking Visual Problem Solving in Generative Models ZODA, Hong Kong University of Science and Technology, University of Chinese Academy of Sciences · 2609.35504 · ▲ 15
- Program-Verified Self-Evolution for Vision-Language Models LGAI Research, Mohamed bin Zayed University of Artificial Intelligence, Australian National University · 2609.33855 · ▲ 14
- Residual Transferability in Neural Image Watermarking National University of Singapore · 2609.32241 · ▲ 12
- InfiniHand: Streaming World-Space Hand Motion Estimation from Egocentric Video Shanghai AI Laboratory, Shanghai Jiao Tong University, University of Science and Technology of China · 2609.35743 · ▲ 9
- Fewer Tokens, More Self-Teaching: On-Policy Self-Distillation for Extreme Visual Token Reduction Shanghai Jiao Tong University, Xi'an Jiaotong University, University of Cambridge · 2609.32353 · ▲ 8
- BaRe-Mem: Bayesian Reliability Memory for Robust and Adaptive Agent Consultation Nanyang Technological University, Peking University · 2609.35551 · ▲ 8
- GeoVerse: World-Consistent Novel View Synthesis in Geometric Latent Space Shanghai Jiao Tong University, Shanghai AI Laboratory, Chinese University of Hong Kong · 2609.35734 · ▲ 8
- WaveFront Decoding: Parallelized Self-Speculative Decoding for Looped Language Models Seoul National University · 2609.23033 · ▲ 7
- Relic: From Multi-Agent Collaboration to Persistent Organizational Capability University of Illinois Urbana-Champaign, Harvey Mudd College, Yale University · 2609.32965 · ▲ 7
- Draft-KV: Learning Useful Latent Communication Between Language Models City University of Hong Kong, University of Science and Technology of China, University of Electronic Science and Technology of China · 2609.34754 · ▲ 7
- ColNanoVDR: Document-Free Query Distillation for Multi-Vector Visual Document Retrieval via Optimal Transport Aalto University, Independent researchers · 2609.34899 · ▲ 7
- RenderRank: Learning to Rerank Text with Compressed Visual Tokens Korea University, Sookmyung Women's University · 2609.35069 · ▲ 7
- Rethinking Automated Voice Similarity by Shifting from EER to Embedding Geometry National Taiwan University, NVIDIA · 2609.33999 · ▲ 6
- SentZero: An Enhanced Sentence-Centric Vision-Language Pretraining for Multi-Task Zero-Shot Chest X-Ray Analysis KAIST · 2609.34479 · ▲ 6
- Can We Trust the Teacher? Decoupled Credit Direction-Magnitude for Self-Distillation Adelaide University, University of North Texas, Curtin University · 2609.34848 · ▲ 6
- Routing Drift Alone Does Not Diagnose Failure in Merged MoE LLMs Hong Kong Polytechnic University, InfiX AI · 2609.32821 · ▲ 5
- KernelZero: Co-Evolving Proposer and Coder for Continuously Improved GPU Kernel Generation Institute of Computing Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Institute of Software, Chinese Academy of Sciences · 2609.33074 · ▲ 5
- DroneWAM: Efficient World Action Model for Drone Visual Navigation Hohai University, Shanghai Jiao Tong University, Nanjing University · 2609.33148 · ▲ 5
- SMAT: Simple and Efficient Merge-Aware Training Hong Kong Polytechnic University, Hong Kong University of Science and Technology, Guangzhou, Chinese University of Hong Kong · 2609.33437 · ▲ 5
- Measuring Collapse and Correction in Homogeneous-Panel LLM Debate Nanyang Technological University · 2609.35279 · ▲ 5
- Not All Objectives Are Born Equal: Priority-Constrained Descent for Hierarchical Multi-Objective Optimization American University of Sharjah, Massachusetts Institute of Technology · 2606.29521 · ▲ 4
- G^2PTQ: Improving LLM Post-Training Quantization with Generalized Gradient Compensation ZTE Corporation, Chinese University of Hong Kong, Northwestern Polytechnical University · 2609.31009 · ▲ 4
- VGGT-Diff: Visual Geometry Meets Diffusion for Sparse-View Novel View Synthesis XPeng Motors, Chinese University of Hong Kong, Tsinghua University · 2609.33253 · ▲ 4
- Beyond Timestamps: Decision-Aligned On-Policy Distillation for Long-Horizon Agents Beihang University, Tsinghua University · 2609.33391 · ▲ 4
- Approximating Softmax in Pretrained LLMs: Model Sensitivity and Kernel Acceleration University of Illinois Urbana-Champaign · 2609.33586 · ▲ 4
- PReCache: Efficient KV Cache Sharing for Multi-LoRA Agents via Low-Rank Precomputation and Neutral Reconstruction Seoul National University · 2609.34054 · ▲ 4
- KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems Seoul National University · 2609.34060 · ▲ 4
- When Do Model Internals Help? Exploring the Role of Representation Engineering in LLM Safety Peking University · 2609.34771 · ▲ 4
- FactorEngram: Factorized N-gram Memory with Basis-Level Gating for Language Models Baidu, Nanyang Technological University · 2609.35578 · ▲ 4
- Reinforcing Agentic Creativity in Scientific Ideation with Night Science University of Illinois Urbana-Champaign, Microsoft · 2609.35706 · ▲ 4
- TokenCast: Forecasting Token Consumption During LLM Agent Execution Sun Yat-Sen University, Rensselaer Polytechnic Institute, Southeast University · 2609.35760 · ▲ 4
- WhiteMatter: All-to-All Cross-Layer Connections via KV Source Mixing University of Southern California · 2608.18486 · ▲ 3
- Rolling-WAM: World Action Models with Rolling Imagination University of Southern California, Brown University, Fudan University · 2609.30247 · ▲ 3
- NVAlign: Direct-Gradient Optimization for Non-Verbal Control in Continuous Autoregressive Flow Matching Text-to-Speech Bland AI · 2609.31892 · ▲ 3
- Playing to Par: Reinforcement Learning for Provably Optimal Quadrilateral Block Decompositions University of California, Berkeley · 2609.32146 · ▲ 3
- Adaptive Consistency Graph for Long-Horizon Agents Beihang University, Qi-Anxin Group · 2609.32754 · ▲ 3
- Allspark: Weak to Strong Transfer via Alternating Chain of Thought University of Texas at Austin, Together AI, Microsoft · 2609.32913 · ▲ 3
- DISCO: Distributed Long Context Scaling with Grounding-Reasoning Disaggregation National University of Singapore, Adobe · 2609.33485 · ▲ 3
- What masking geometry works best for EEG foundation models? Radboud University, Yneuro, University of California San Diego · 2609.33487 · ▲ 3
- Pruned CTC for Memory-Efficient Large-Vocabulary ASR Training Shanghai Jiao Tong University, Alibaba, Tsinghua University · 2609.33645 · ▲ 3
- Who Gets a Token, and What Does It Carry? Unequal Name Support and Concept Access in Large Language Models University of Alberta · 2609.34065 · ▲ 3
- Geometry as Address: Routing Attention to Visual Memory for Long-Horizon Camera-Controlled Video Generation Zhejiang University, Hong Kong University of Science and Technology, Guangzhou, Lightspeed · 2609.34722 · ▲ 3
- On-Policy Self-Distillation for Multi-Turn Image Editing King Abdullah University of Science and Technology, Krea AI · 2609.35611 · ▲ 3
- EvolvingAvatar: Interactive 3D Head Generation That Adapts as Conversations Unfold Hefei University of Technology, Shanghai Jiao Tong University, Hefei Comprehensive National Science Center · 2609.35616 · ▲ 3
- Distillation Defenses Easily Break After Reinforcement Learning University of Oxford, ELLIS Institute Tübingen, Max Planck Institute for Intelligent Systems · 2609.35699 · ▲ 3
- Safe Error Correction for Language Models: Frozen-Base Adjustment with Capability Preservation Eulogik · 2609.16145 · ▲ 2
- NanoForecast v0.5: Competitive Time Series Forecasting Through Training Pipeline Optimization Eulogik · 2609.31669 · ▲ 2
- SkillDRE: Dual-Stage Red-Team Evolution of Agent Skills via Pre-Execution and Runtime Feedback Beijing University of Posts and Telecommunications, North China Electric Power University, Chongqing University of Posts and Telecommunications · 2609.32400 · ▲ 2
- SCOPD: Sparse-Context On-Policy Self-Distillation for Efficient Vision-Language Models University of Toronto, Vector Institute, Samsung · 2609.34044 · ▲ 2
- AnswerMap: Faithful Spatial Interpretability of VLMs from Answer Posteriors King Abdullah University of Science and Technology, Edge Hill University · 2609.35247 · ▲ 2
- Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning Peking University, Shenzhen University · 2609.35472 · ▲ 2
- LeRF: Learning Reference Coordinate Frames for Perspective Taking Reasoning University of Illinois Urbana-Champaign, Shanghai Jiao Tong University, University of California, Merced · 2609.36219 · ▲ 2
Monday, September 28, 2026 · 33
- FuseReg: Regularizing Layer Fusion Mitigates the Reconstruction-Generation Gap in Representation Autoencoders University of Southern California, Brown University, Rice University · 2609.31620 · ▲ 157
- Disaggregated Quantization: Specializing LLM Prefill and Decode NVIDIA, Institute of Science and Technology Austria · 2609.26333 · ▲ 90
- RayOrch: Programming and Executing Lineage-Controlled Multi-Grain Dataflows for Foundation-Model Data Preparation Peking University, Hong Kong University of Science and Technology, University of Cambridge · 2609.18703 · ▲ 54
- TimeEvo: Failure-Driven Self-Evolution of a Time Series Agent University of Illinois Chicago, Visa Research, University of Florida · 2609.27277 · ▲ 31
- Block Sparse Attention with Log-Linear Complexity Shanghai Jiao Tong University, Shanghai Innovation Institute, ByteDance · 2609.31093 · ▲ 30
- LastOPD: Taming Collapse in Latent On-Policy Distillation University of Illinois Chicago, Case Western Reserve University, University of Florida · 2609.28845 · ▲ 28
- InternW0-Δ: A World Action Model Bridging Predictive Dynamics and Actions with 20K+ Hours of Open Data Shanghai AI Laboratory · 2609.31394 · ▲ 27
- Tactile-JEPA: Topology-Aware Self-Supervised Representation Learning for Distributed Tactile Sensors Sber AI, Skolkovo Institute of Science and Technology, HSE University · 2609.24385 · ▲ 22
- FoMo: Forking Moment in Generative Trajectory as a Perceptual Distance Seoul National University · 2609.25716 · ▲ 22
- IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking National Payments Corporation of India · 2609.29167 · ▲ 20
- D-JEPA: A Decision-Aligned Latent World Model Hong Kong University of Science and Technology, Guangzhou, Boston University, Shanghai Jiao Tong University · 2609.24749 · ▲ 17
- Jev in the Wild: A Data-Driven Analysis of the Jev Model's Functionality, Applications and Ecosystem Sun Yat-sen University, Chinese University of Hong Kong · 2609.30216 · ▲ 15
- Enhancing Photogrammetric Digital Surface Models with Pretrained Diffusion Models and Multimodal Conditioning Thales, Centre National d'Études Spatiales · 2609.31199 · ▲ 14
- Morphometric Imitation: From Morphology and Contact Aware Hand Retargeting to Sim-to-Real Visuomotor Policy University of California, Berkeley · 2609.28660 · ▲ 13
- TrackEverything: Long Horizon Dense Tracking via De-Duplicating 3D Scene Representations Carnegie Mellon University, Meta · 2609.30222 · ▲ 13
- SLCA-GRPO: Resolving Cross-Segment Credit Misattribution in Tool-Calling RL Peking University, Shenzhen University, Tencent · 2609.29050 · ▲ 12
- Game Arena: Strategic LLM Evaluation in Competitive Environments arXiv · 2609.31473 · ▲ 12
- Depth-adaptive Inference of Looped Language Models via Continuous Depth Batching Technical University of Munich, Imperial College London, Munich Center for Machine Learning · 2608.09444 · ▲ 11
- TRACE: Temporal Audit and Condition-aware Evaluation of Streaming Video Understanding Om AI Research · 2609.30670 · ▲ 11
- AgentWorld: Benchmarking Long-Horizon Collaboration of Multi-agent LLMs OpenAgents, Columbia University, University of Pennsylvania · 2609.31590 · ▲ 11
- Do Implicit Personalization and Explicit Styles Conflict? PsPLUG: A Lightweight Plug-in for Balancing Personalization and Style in Customized LLMs University of California, Irvine, Independent researchers, ByteDance · 2601.06362 · ▲ 10
- LightMIS: Ultra-Lightweight Medical Image Segmentation Without a Stage-Wise Decoder Alexandru Ioan Cuza University of Iasi, University of Wurzburg · 2609.28327 · ▲ 10
- VLA-Precision: Asymmetric Co-Bootstrapping for Efficient Real-World Online RL of Vision-Language-Action Models University of Science and Technology of China, State Key Laboratory of Precision and Intelligent Chemistry, Beihang University · 2609.04355 · ▲ 9
- SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance Virginia Tech, University of Wisconsin Madison, Dartmouth College · 2609.30192 · ▲ 8
- CARD: Cluster-level Adaptation with Reward-guided Decoding for Personalized Text Generation University of California, Irvine, Independent researchers, University of Amsterdam · 2601.06352 · ▲ 7
- Not All Ranks Are Equal: Budget-Aware LoRA Merging Across Tasks Samsung · 2609.22237 · ▲ 7
- BoundInk: Boundary-Aware Online Handwriting Generation Sungkyunkwan University · 2604.02103 · ▲ 6
- Paragraph Boundaries Are Not White Space:Compression Depth as the Signature of Hierarchical Structure arXiv · 2609.23551 · ▲ 6
- MOPD-Router: Rethinking Teacher Routing in Multi-Teacher On-Policy Distillation Shanghai Jiao Tong University, Alibaba, Shanghai Innovation Institute · 2609.30837 · ▲ 6
- Softmax Reparameterization for Output-Head Quantization Adobe · 2609.31291 · ▲ 6
- Evidence-Grounded Auditing of Identification Assumptions in Climate-Policy Causal Evaluations Autonomous University of Madrid, Spanish National Research Council · 2609.30867 · ▲ 5
- CodeGraph: Open-Taxonomy Knowledge Graph for Source Code with Wikidata Grounding University of Bologna, Télécom Paris, Sant'Anna School of Advanced Studies · 2609.29474 · ▲ 4
- ZooWork-ShopRanker: An Open, Preference-Aligned E-Commerce Reranker ZooWork Team · 2609.31002 · ▲ 3
Friday, September 25, 2026 · 22
- Training Object Permanence in World Models University of Southern California, Carnegie Mellon University, University of Michigan · 2609.28654 · ▲ 212 · Deep dive
- Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs arXiv · 2609.29845 · ▲ 81
- WanPE: Towards Cinematic Prompt Enhancement for Modern Text-to-Video Generation Nanjing University, Alibaba, University of Science and Technology of China · 2609.30221 · ▲ 37
- OmniEcho: Spatial Audio Understanding for Embodied Agents Peking University, Alibaba, Tsinghua University · 2609.23407 · ▲ 26
- Agent-Editing World Model: Rethinking World Modeling for LLM Agents Renmin University of China · 2609.28416 · ▲ 24
- Rufus-Air: An Open LLM Post-Training Recipe Amazon · 2609.29421 · ▲ 22
- ExplorationBench: Measuring AI Systems' Exploration in Verifiable Alien Worlds Fudan University, Tencent, Tsinghua University · 2609.30199 · ▲ 19
- Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents Alibaba · 2609.29892 · ▲ 18
- Learning to Discover Interesting Mathematics Meta, New York University, CERMICS at the École nationale des ponts et chaussées, Institut Polytechnique de Paris · 2609.28603 · ▲ 14
- PUBG Ally: A Conversational Embodied Agent as an AI Teammate Krafton AI · 2609.29837 · ▲ 14
- Parts-of-Speech as Emergent Categories in SAE Latent Space University of Pisa · 2609.29362 · ▲ 13
- Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures Griffith University, Nanyang Technological University, UNSW Sydney · 2609.29429 · ▲ 13
- Coding Agents for Generalized Task and Motion Planning Problems Princeton University, Fondazione Bruno Kessler, Carnegie Mellon University · 2609.30233 · ▲ 12
- IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis Zhejiang University, Tencent · 2609.29444 · ▲ 9
- AgentKernel: The Trust-Native Agentic Operating System DeepKernel Lab · 2609.29647 · ▲ 9
- Neural Spectral Capacity: Measuring and Designing Architectures from Network Specification Alone City University of Hong Kong · 2609.23087 · ▲ 8
- RGBD20K: A Large-Scale Benchmark for RGB-D Semantic Segmentation University of North Texas · 2609.29028 · ▲ 8
- World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal Hong Kong University of Science and Technology, Guangzhou, Chinese University of Hong Kong, Knowin AI · 2609.29964 · ▲ 8
- DeltaWAM: Delta World Action Models for Bimanual Manipulation Peking University, AI2 Robotics · 2609.28811 · ▲ 6
- AV-GRPO: Modality-Anchored Decoupling Diffusion Reinforcement Learning for Joint Audio-Video Generation Shanghai AI Laboratory, Hong Kong Polytechnic University, National University of Singapore · 2609.29816 · ▲ 5
- Rate-distortion optimization for full-reference image quality metrics via stochastic Hessian estimates University of Southern California · 2609.30077 · ▲ 4
- ViRDM: Taming Representation Distribution Matching for Few-Step Causal Video Generation Northeastern University, Adobe · 2609.28923 · ▲ 3
Thursday, September 24, 2026 · 29
- SpeakerMem-R1: Speaker-Centered Dual-Track Memory for Multi-Party Dialogue Zhejiang University · 2609.26780 · ▲ 85
- The Past Frames the Future: Memory for Autoregressive Video Generation Hong Kong University of Science and Technology, City University of Hong Kong, Fudan University · 2609.28466 · ▲ 47
- Spatial-Interactor: Learning Spatial Reasoning through Interaction with the Observable Physical World Zhejiang University, SAP · 2609.23038 · ▲ 46
- HappyWorld-Bench Alibaba, Beijing Institute for General Artificial Intelligence, Tsinghua University · 2609.24308 · ▲ 43
- Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM Agents Salesforce · 2609.27334 · ▲ 36
- RewardVerse: Rubric-Guided Policy Optimization for Video Reward Modeling Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Hong Kong University of Science and Technology · 2609.22947 · ▲ 24
- Schrödinger's Code Repository: Have LLMs Learned SWE-bench or Memorized It? Shanghai Jiao Tong University, Xi'an Jiaotong University, East China Normal University · 2609.27891 · ▲ 20
- Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms Georgia Institute of Technology, Alibaba · 2609.27321 · ▲ 17
- PACT: From Credit Assignment to Critic Alignment AllSpark Team · 2609.26355 · ▲ 16
- Capable yet Parsimonious: Extracting and Characterizing Hidden Chain-of-Thought in Frontier Models Aalborg University, Seafill · 2609.26637 · ▲ 14
- Hunyuan-A13B Technical Report Tencent · 2609.27284 · ▲ 14
- GeoPair: Geometry-Preserving Cross-Layer Factorization for Training-Free Transformer Compression MWS AI, ITMO University · 2609.25963 · ▲ 13
- PackLab: A Comprehensive Framework for Developing, Training, and Evaluating MLLMs in Robotic Bin Packing Chinese University of Hong Kong, M-A-P · 2609.23784 · ▲ 11
- WhatWorkedBench: Benchmarking Experimental Understanding in AI Agents Carnegie Mellon University, Tsinghua University · 2609.27490 · ▲ 10
- InternW0: A Foundational Physical World Model for Efficient Real-World Interactions Shanghai AI Laboratory · 2609.27656 · ▲ 10
- MemBodied: Recurrent Associative Memory for Vision-Language-Action Models Nanyang Technological University, Griffin Labs, École Centrale de Lyon · 2609.28256 · ▲ 10
- All modalities are equal, but video is more equal: Closing the Cross-Attention Gap in Joint Video Generation Bar-Ilan University, NVIDIA · 2609.27901 · ▲ 8
- On the Diffusibility of High-Dimensional Latents Cornell University, Adobe, Virginia Tech · 2609.28473 · ▲ 7
- Self-Organizing Agent Teams Learn to Reason Together Stanford University, Together AI, Emory University · 2609.22682 · ▲ 6
- MemoryAthena: Adaptive Routing over Latent and Generated Memories ELLIS Institute of Finland, University of Turku, University of Technology Sydney · 2609.25853 · ▲ 4
- FLEET: From Logits Entropy to Enhanced Trajectories in Text Generation Kharkiv National University of Radio Electronics · 2609.27657 · ▲ 4
- Six Layers Less: Encoder Pruning for Whisper with Label-Free Recovery Technical University of Denmark, Laerdal Medical · 2609.27980 · ▲ 4
- The Linear Representation Hypothesis Needs a Group Action Independent researchers, Chinese University of Hong Kong · 2609.27158 · ▲ 3
- EmbodiedSWE: Coding Agents for Long Horizon Dexterous Robotics ByteDance, Yale University, Princeton University · 2609.27308 · ▲ 3
- StudentBench: AI and human tutoring yield equivalent GRE learning gains arXiv · 2609.28470 · ▲ 3
- Uranus: Building the Next-Generation Simulation Infrastructure for Embodied AI D-Robotics · 2609.24815 · ▲ 2
- X-Planner: Event-Structured Task Planning for Embodied Intelligence X Square Robot · 2609.25187 · ▲ 2
- Calibration as a First-Class Criterion in LLM Evaluation Johannes Gutenberg University Mainz, University of Colorado Boulder · 2609.26489 · ▲ 2
- Knowledge Pull Requests for Continual Document Authoring Johns Hopkins University · 2609.26634 · ▲ 2
Wednesday, September 23, 2026 · 24
- The Tasteful Agent: Measuring and Improving Taste in Long-Horizon Tasks City University of Hong Kong, Independent researchers, Microsoft · 2609.25804 · ▲ 156 · Deep dive
- RULER: Instance-aware Rubric Rewards for SVG Generation Ant Group, Hong Kong University of Science and Technology, Guangzhou, Independent researchers · 2609.25270 · ▲ 100
- Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings Alibaba · 2609.25165 · ▲ 72
- GAE: Learning a Geometry-Native Latent Space for 3D-Consistent World Generation Hong Kong University of Science and Technology, Tencent, University of Hong Kong · 2609.24981 · ▲ 66
- All-in-One Multilingual Scene Text Recognition with Script-aware Mixture-of-Experts Fudan University, Shanghai Key Laboratory of Multimodal Embodied AI, Tencent · 2609.24058 · ▲ 55
- StableVQ: Practical Guidelines for Stable Vector-Quantized Tokenizer Training Huazhong University of Science and Technology, Kuaishou, South China Normal University · 2609.26774 · ▲ 54
- Bellman Policy Optimization Apodex, Princeton University · 2609.15987 · ▲ 48
- JEV-as-a-Judge: Accept When Confident, Escalate When Unsure Carnegie Mellon University · 2609.26550 · ▲ 41
- Circuit Hypernetworks for Quantum-Augmented Diffusion Language Models University of Montreal, Mila, McGill University · 2609.24657 · ▲ 40
- Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs Mohamed bin Zayed University of Artificial Intelligence · 2609.26796 · ▲ 36
- From Pattern Recognizers to Personalized Companions: A Survey of Large Language Models in Mental Health arXiv · 2609.25186 · ▲ 31
- Lean Pool: An AI-Maintained Archive of Formalized Mathematics University of Washington · 2609.25199 · ▲ 27
- Recursive self-improvement of AI research agents Weco AI · 2609.26457 · ▲ 27
- Agensh: Scaling Organizational Intelligence to 1,024 Agents Microsoft · 2609.26781 · ▲ 25
- Emergent Collusion in Long-Horizon LLM Agent Interaction Stanford University, Georgia Institute of Technology · 2609.24967 · ▲ 19
- Embedding Physics Priors in Robot Learning: A Survey Technical University of Munich, Technical University of Darmstadt, University of Trento · 2609.22319 · ▲ 17
- Tri-PvP: Exposing Modality Bias in Omni-Modal Large Language Models through Perceptual-Propositional Evidence Conflicts National Taiwan University, NTU Artificial Intelligence Center of Research Excellence · 2609.06011 · ▲ 15
- LatentPort: Beyond KV Cache - Cross-Model Transfer of Recurrent Memory in Hybrid Language Models: A 4B-to-9B Hybrid-State Handoff Without Target Prefix Replay arXiv · 2609.25053 · ▲ 15
- ImIR: Image-Instruction Tuning for All-in-One Image Restoration Codeway AI Research · 2609.25267 · ▲ 13
- RoboFollow: Unveiling the Instruction Following Mirage in Embodied Agents Shanghai Jiao Tong University, Anyverse Dynamics · 2609.25636 · ▲ 11
- Blaming Across the Aisle: Political Contrasting and Blame Attribution in the Danish Parliament Aarhus University, Center for Humanities Computing · 2609.26346 · ▲ 11
- ALPINE: Adaptive Localization for Parameter- and Sample-Efficient Few-Shot Learning Independent researchers · 2609.22323 · ▲ 10
- Geometric and Semantic Coupling for Interaction Understanding in 3D Scenes National University of Singapore, Hong Kong Polytechnic University · 2609.25247 · ▲ 10
- HARMONY: Hierarchical Agentic Reasoning for MONocular Image-to-Scene Synthesis University of Pennsylvania · 2609.26793 · ▲ 6
Tuesday, September 22, 2026 · 34
- Realtime-Venus: A full-duplex interaction system with asynchronous delegation Ant Group, Tsinghua University · 2609.13814 · ▲ 210 · Deep dive
- RRSI: Regularized Recursive Self-Improvement of Agent Harnesses Google, University of North Carolina at Chapel Hill, Stanford University · 2609.24972 · ▲ 200 · Deep dive
- OmniEdu: Open Foundation Models for Learning and Teaching Peking University, University of Chinese Academy of Sciences, Zhongguancun Academy · 2609.23088 · ▲ 167 · Deep dive
- WorldCrafter: Consistent Video World Model with Implicit 3D-aware Memory Tencent, Peking University · 2609.24984 · ▲ 146
- GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay Tencent, Great Bay University, National University of Singapore · 2609.25001 · ▲ 130
- Transferring the Intelligence of VLMs to Robotic Control Tsinghua University, Tencent · 2609.22966 · ▲ 116
- Grounded Action Model: 3D Grounding as a Foundation for Robotics Northwestern University, University of Washington, National University of Singapore · 2609.23863 · ▲ 85
- Document Retrieval-Aware Chunking (D-RAC): Universal Retrieval-Aware Ingestion of Enterprise Documents via PDF Normalization and Multimodal Markdown Conversion Yellow.ai · 2609.24220 · ▲ 65
- VideoGen-Agent: Reinforcing Video Generation Agents Princeton University, Stanford University, University of California, Davis · 2609.24997 · ▲ 65
- onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction StepFun, Xiamen University · 2609.24983 · ▲ 54
- One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering Agents Alibaba · 2609.23377 · ▲ 48
- Harness-Zero: Harness Distillation via Agent-as-Harness Peking University, Google, Hong Kong University of Science and Technology · 2609.24974 · ▲ 35
- Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents University of Texas at Dallas · 2609.23986 · ▲ 26
- Why Do Video Diffusion Models Violate Physics? Unveiling the Flaws in Attention Mechanisms Beijing University of Posts and Telecommunications · 2609.23658 · ▲ 25
- CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies Dalian University of Technology · 2609.24118 · ▲ 24
- Deep Persona: A Psychologically Grounded Architecture and Evaluation Framework for Role-Playing Agents and Simulations University of Haifa, Bar-Ilan University · 2609.22255 · ▲ 22
- HuRo: Robotizing Human Videos for Scalable VLA Pretraining RL World, Yonsei University · 2609.10706 · ▲ 16
- 1% of Tokens Can Be Enough: On Gradient Estimation in On-Policy Distillation Mohamed bin Zayed University of Artificial Intelligence, Ant Group · 2609.24432 · ▲ 16
- Towards Full Pipeline FP8 Reinforcement Learning for LLMs University of Wisconsin-Madison, ByteDance · 2609.22870 · ▲ 15
- ShieldVLA: Feasibility-Aware Safety Alignment for Vision-Language-Action Models Indian Institute of Science, Microsoft · 2609.13231 · ▲ 13
- The Functionalizer: Lossless Functional Decomposition for Subword Tokenization Massachusetts Institute of Technology · 2609.15991 · ▲ 13
- ACLArena: Agent Continue Learning in Multi-stage Post-training University of California, Los Angeles, University of California, Santa Cruz · 2609.23989 · ▲ 12
- Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies University of Wisconsin-Madison, University of Illinois Urbana-Champaign · 2609.24682 · ▲ 11
- Streaming Video Editing with Easy Adaptation Shanghai Jiao Tong University · 2609.24788 · ▲ 10
- Complex KDA: Understanding and Enhancing the Expressivity of Kimi Delta Attention University of Freiburg, Microsoft, University of Tübingen · 2609.24797 · ▲ 10
- A Lie Detector Test for Language Models: Reading Knowledge a Model Won't Reveal StackOne Technologies · 2609.21996 · ▲ 9
- SkillSpec: Intent-Masked Specification Reasoning for Agent Skill Correctness Beihang University · 2609.06052 · ▲ 8
- Measuring the Checker: Mutation Analysis for GPU-Kernel Benchmark Oracles National University of Singapore, Nanyang Technological University, VinUniversity · 2609.22220 · ▲ 7
- The information geometry of large language models is shared, learned, and controllable University College London, London Centre for Nanotechnology · 2609.11063 · ▲ 5
- Prediction-Powered Smoothing and Validation for Disaggregated AI Evaluation University of California, Santa Cruz · 2609.20758 · ▲ 4
- Mira-Scene: Pixel-Aligned Layouts for Generative 3D Scene University of Hong Kong, VAST · 2609.23796 · ▲ 4
- EDGEGEN: Improving Tool-Calling Agents Beyond Happy Paths with Synthetic Edge Case Generation SAP · 2609.24115 · ▲ 4
- UltraTex: Unleashing 2K Multi-View Diffusion for 3D Texturing Jilin University, Shanghai Innovation Institute, University of Hong Kong · 2609.23169 · ▲ 3
- TAPe+ML: A Compact Structured Representation for Multi-Task Computer Vision Comexp Research Lab · 2609.20869 · ▲ 2
Monday, September 21, 2026 · 31
- EvoOntology: A Self-Evolving Ontology Layer for Data Agents Renmin University of China · 2609.15779 · ▲ 150
- OmniVChat: Synthesizing, Benchmarking, and Training for Native Audio-Visual Dialogue Chinese University of Hong Kong, Alibaba, Shanghai Jiao Tong University · 2609.21465 · ▲ 143
- CodeMidas: Scaling Agentic Coding RL Environments from Code Itself Xiaomi, Peking University, University of Hong Kong · 2609.22068 · ▲ 131 · Deep dive
- Grounded Skill Synthesis from Code at Scale for Agentic Intelligence Ant Group · 2609.05571 · ▲ 113
- RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents Alibaba · 2609.22000 · ▲ 78
- Paint-Anything: Unified Any-Color Control for Image Generation and Editing ByteDance, Zhejiang University, Nanjing University · 2609.20816 · ▲ 54
- IntBMoE: Integrating Block-Level Conditioning into Expert Composition for Full-Participation Mixture-of-Experts Alibaba · 2609.21346 · ▲ 44
- OmniVBench: A Benchmark and Large-Scale Dataset for Omni Reference-to-Video Generation Tencent, Hong Kong University of Science and Technology, Guangzhou · 2609.22069 · ▲ 35
- Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic Design Adobe, Brown University · 2609.22086 · ▲ 33
- MintAct: A Unified Visual Agent for Digital Environments Apple · 2609.22083 · ▲ 32
- BI-Agent and BI-Bench: Towards Automating End-to-End Business Intelligence University of Illinois Urbana-Champaign, Microsoft · 2609.20886 · ▲ 28
- GraphSkillEvo: Evolutionary Optimization of Graph-Structured Agent Skills City University of Hong Kong, National University of Singapore, Southern University of Science and Technology · 2609.21749 · ▲ 18
- DeformSmith: Physics Harness-Guided Hierarchical Generation of Deformable Assets for Robot Manipulation Nankai University, Rightly Robotics · 2609.18620 · ▲ 14
- Calibrating Teacher--Student Discrepancy for On-Policy Distillation Nanjing University, Southeast University, Nanjing University of Posts and Telecommunications · 2609.21619 · ▲ 13
- From Pretraining to Proficiency: Real-World Subtask RL for Long-Horizon Manipulation with Minimal Human Intervention University of Texas at Austin, Autel US, University of California, Berkeley · 2609.21788 · ▲ 13
- Gricea: An Open Science Platform for Conversational AI Research Johns Hopkins University · 2609.22039 · ▲ 13
- MoME: Mixture-of-Memory Embeddings for Context-Aware Sparse Lookup University of British Columbia, Vector Institute · 2609.15126 · ▲ 12
- CADWorld: Computer-Use Benchmark for Long-Horizon Computer-Aided Design Georgia Institute of Technology, North Carolina State University, Marquette University · 2609.16251 · ▲ 12
- MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads Shenzhen University of Advanced Technology · 2609.09206 · ▲ 11
- Refinement Is Inherently Editable: Training-Free Prompt-to-Prompt Image Editing with Generative Refinement Network City University of Hong Kong, Dongguan, City University of Hong Kong, Mohamed bin Zayed University of Artificial Intelligence · 2609.20633 · ▲ 11
- When AI Reviews Train AI Reviewers: Scientific-Judgment Collapse and Mitigation University of Maryland · 2609.20942 · ▲ 9
- TeleAntiFraud 2.0: A Refreshable, Profile-Grounded, and Audio-Based Benchmark for Telecom Fraud Detection People's Public Security University of China, JD.com, Ant Group · 2609.18748 · ▲ 8
- Learning Foresight without Explicit Trajectories for 3D Diffusion Policies Dalian University of Technology · 2609.20669 · ▲ 8
- FRAUDSkill: Structured Frozen-Weight Skill Optimization for Audio Anti-Fraud Detection People's Public Security University of China, JD.com, Meta · 2609.18766 · ▲ 6
- SiliconBench: Speed, Memory, and Fidelity for LLM Serving on Unified-Memory Desktops Pennsylvania State University, University of Illinois Urbana-Champaign, EXO Labs · 2609.19169 · ▲ 6
- Training-Adaptive Convolutional Sparse Coding via Information Bottleneck for Robust Visual Representation Shenzhen University of Advanced Technology, University of Electronic Science and Technology of China, Shandong University · 2609.19122 · ▲ 5
- Geometry of Values: Task Vector Composition for Ethical Preference Alignment in Language Models Mohamed bin Zayed University of Artificial Intelligence · 2609.21094 · ▲ 5
- SteerDuplex: Steerable Duplex Speech Dialogue Models Scale AI, University of Maryland · 2609.12623 · ▲ 4
- GAVEL: Graph World Models for Verified and Efficient Long-Horizon LLM Task Planning arXiv · 2609.19315 · ▲ 4
- APort Vault: Benchmarking AI Agent Payment Authorization with the Open Agent Passport APort Technologies · 2609.22076 · ▲ 4
- Retention-Constrained Post-Training Quantization of Cellpose-SAM for Stem Cell Microscopy Capicú Technologies · 2609.21038 · ▲ 3
Friday, September 18, 2026 · 24
- DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression DeepSeek · 2609.19969 · ▲ 178 · Deep dive
- SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness NVIDIA, NTU, Massachusetts Institute of Technology · 2609.20519 · ▲ 130
- Can MiniMax-H3 Reason About the Physical World? An Evaluation of Omni-Modal Generative Model National University of Singapore, Fudan University, Tencent · 2609.18323 · ▲ 113
- When EOS Tokens Disagree: Understanding Length Inflation in On-Policy Distillation University of North Carolina at Chapel Hill, Brigham Young University, Microsoft · 2609.20511 · ▲ 107
- An Empirical Study of Harness Design for Coding Agents University of Massachusetts Amherst, Emory University, University of North Carolina at Charlotte · 2609.20804 · ▲ 85
- JEPA-Anything: Learning Predictive Models across Different Worlds arXiv · 2609.20800 · ▲ 70
- Verifiable Social Reasoning for LLM Assistants Google, Hebrew University, University of Cambridge · 2609.17496 · ▲ 65
- Self-Evolving Search Index Yonsei University, Samsung, University of California, Irvine · 2609.19656 · ▲ 60
- RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning Zhejiang University, Alibaba · 2609.20784 · ▲ 56
- When2Think: Learning Difficulty-Aware Length Control for Efficient Hybrid Reasoning Models Sungkyunkwan University, Microsoft · 2609.19671 · ▲ 50
- Video DeltaNet: A Video-Native Hybrid Attention for Livestream Video Generation University of California, Berkeley, Impossible Inc, University of Texas at Austin · 2609.20744 · ▲ 50
- RiskChainBench: A Benchmark for Obfuscated Platform Message Restoration and Evidence-Grounded Web Investigation Baidu, SmartFlow AI, People's Public Security University of China · 2609.16900 · ▲ 47
- WeVisDoc: From Coverage to Capability for Robust End-to-End Document Parsing Tencent · 2609.20423 · ▲ 47
- Reflect, Revise, Reuse: Training-Free Skill Evolution for GUI Agents Zhejiang University, University of Electronic Science and Technology of China · 2609.17653 · ▲ 45
- UFO: Chain-of-Evaluation for Omni-Condition Alignment in Multi-Modal Image Generation University of Science and Technology of China, Independent researchers, Shenzhen University · 2609.12397 · ▲ 43
- Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL University of Maryland, Amazon · 2609.20715 · ▲ 42
- VABench: Measuring Embodied Spatial Intelligence through Visual Demonstrations, Active Perception, and Metric Control Dalian University of Technology, Nanyang Technological University · 2609.19554 · ▲ 41
- Region-Level Policy Optimization for Fine-grained MLLM Perception University of Sydney, City University of Hong Kong · 2609.19745 · ▲ 41
- PACT: Can Enterprise AI Assistants Be Trusted Under Pressure? Georgia Institute of Technology, Decagon AI, Baseten · 2609.18605 · ▲ 40
- FAMOS: Feed-Forward 3D Articulation Modeling from Sparse Observations Stanford University, ETH Zurich · 2609.20817 · ▲ 38
- Srijika: OpenType-Layout-Reusing Font Restyling for Nine Indic Scripts Loopdesk Technologies · 2609.05661 · ▲ 37
- Sample Count Is Not Enough: Candidate-Generation Strategy Shapes the Energy and Performance of LLM Test-Time Scaling Iowa State University · 2609.19499 · ▲ 37
- What Does Privileged Information Add to On-Policy Self-Distillation? National University of Singapore · 2609.20612 · ▲ 36
- VākQA: A Benchmark and Evaluation Study for Telugu Spoken Factoid Question Answering International Institute of Information Technology, Hyderabad, Brno University of Technology · 2609.19879 · ▲ 31
Thursday, September 17, 2026 · 23
- LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence Stable AI, Tsinghua University · 2609.17488 · ▲ 743 · Deep dive
- ScienceIDE: Turning World's Scientific Codebase into Agent Learnable Environments Aitonomy Foundation, University of Oxford, University of California, Los Angeles · 2609.19134 · ▲ 100
- Rethinking Critic Learning in PPO: Understanding and Mitigating Value Flattening Shanghai Jiao Tong University, Shanghai AI Laboratory, Westlake University · 2609.18708 · ▲ 79
- Confidence Comes from Experience: Experiential Confidence Estimation from Reasoning to Agents University of Cambridge, Google · 2609.17708 · ▲ 72
- ProgramDistill: From Interactive Web Apps to Verifiable Reference-Guided SWE Tasks KAIST, Microsoft · 2609.18805 · ▲ 62
- Agora: Git as Shared Memory for Collective AutoResearch NVIDIA · 2609.18094 · ▲ 57
- VC-Attention: Value Smoothing and Softmax Casting for Low-bit Attention Nunchux AI, Massachusetts Institute of Technology, Carnegie Mellon University · 2609.15810 · ▲ 50
- ActionPiece: Rethinking Action Tokenization for Autoregressive Vision-Language-Action Models Huazhong University of Science and Technology, Zhongguancun Academy, DeepCybo · 2609.18487 · ▲ 47
- Zing-0.5: Toward Playable Worlds with Real-Time Joint Action and Text Control SeedLeap.ai · 2609.17909 · ▲ 44
- EvolveTrade: Experience-Driven Policy Refinement for Self-Evolving LLM Trading Agents KAIST, DeepAuto.ai · 2609.17632 · ▲ 39
- SpectralShift: Effective Context Window Extension of Gated DeltaNet via Spectral Reparameterization Renmin University of China, IQuest Research, Microsoft · 2609.14320 · ▲ 33
- HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses University of California San Diego, Texas A and M University, Carnegie Mellon University · 2609.15938 · ▲ 30
- Gaze as Evidence for Common Grounding: A Cross-Corpus Analysis of MapTask and MUNDEX Utrecht University, Queen Mary University of London · 2609.18011 · ▲ 30
- A Zeroth-Order Paradigm for LLM Preference Alignment University of California, Berkeley, New York University, Alibaba · 2609.19144 · ▲ 29
- EventEgoHands++: Event-based Egocentric 3D Hand Mesh Reconstruction with Real Dataset Keio University, JST Presto · 2609.17189 · ▲ 28
- In-Context Robot Learning with VLM Agents Morphi Robot, Shanghai Innovation Institute, Huazhong University of Science and Technology · 2609.19138 · ▲ 22
- Flattening Every Memory Peak in Long-Context Mixture-of-Experts Training Salesforce · 2609.14306 · ▲ 19
- PANORAMA: Panoptic Grounded Captioning via Mask Proposal Selection Inria, École normale supérieure, CNRS · 2609.19143 · ▲ 17
- The Other Half of the Memory Wall: Serving 35B MoEs from SSD with Trained Routing Prediction AutoArk · 2609.18063 · ▲ 16
- CERA-MoA: Co-Evolving Routing Mechanisms with Continually Learning LLM Agents Tsinghua University, Shanghai Jiao Tong University, Shanghai Qi Zhi Institute · 2609.18779 · ▲ 14
- Fathom: Per-Query Read Depth for Sparse Decoding over Offloaded KV Caches arXiv · 2609.17652 · ▲ 6
- Assessing nnU-Net Generalization across Brain Tumor Populations in BraTS-GoAT 2026 University of Strasbourg, Institut Strauss · 2609.15524 · ▲ 3
- Fingers as Legs: Learning Self-Supported Locomotion and Manipulation with an Anthropomorphic Hand ETH Zurich · 2609.17172 · ▲ 3
Wednesday, September 16, 2026 · 25
- Continual Learning Mechanisms Compose for Long-Horizon Memorization Johns Hopkins University · 2609.06986 · ▲ 373 · Deep dive
- ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement Beihang University, University of Manchester, IQuest Research · 2609.14857 · ▲ 212
- AI for Games in the Foundation Model Era National University of Singapore, Nanyang Technological University · 2609.16679 · ▲ 145
- StepAudio 3 Realtime Technical Report StepFun · 2609.14005 · ▲ 126
- StepAudio 3 Music Technical Report StepFun, ACE, Chinese University of Hong Kong · 2609.16034 · ▲ 108
- Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement Tianjin University, Shanxi University · 2609.13406 · ▲ 82
- ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agents arXiv · 2609.17523 · ▲ 30
- HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness Nanjing University, AgiBot, Tsinghua University · 2609.15195 · ▲ 21
- The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement Shanghai Jiao Tong University, Theseus Labs, Tsinghua University · 2609.11873 · ▲ 17
- FLAT: Resampling Image and Text into 1D Flexible-Length Aligned Transmodal Tokens for Retrieval and Generation Meta · 2609.16591 · ▲ 16
- Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models Shanghai AI Laboratory, Nanjing University, University of British Columbia · 2609.13680 · ▲ 13
- Another Blueprint In The Wall: How to Ask Frontier AI Like a Kid? University of Luxembourg · 2609.14803 · ▲ 12
- ImpossibleRubrics: Stress-Testing Generated Rubrics as Reward Signals National University of Singapore, Peking University, Institute of Automation, Chinese Academy of Sciences · 2609.16816 · ▲ 11
- Modality-Autoregressive World-Action Models Carnegie Mellon University · 2609.17524 · ▲ 11
- PhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion Control University of Pennsylvania, Snap, King Abdullah University of Science and Technology · 2609.17521 · ▲ 10
- Training Specialist Models without Reasoning Trajectories for Domain Expert Distillation University of British Columbia, Shanghai AI Laboratory, University of Helsinki · 2609.13770 · ▲ 9
- Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States University of California San Diego, KU Leuven, University of Illinois Chicago · 2609.15972 · ▲ 9
- Disentangling Representation Evolution in Transformers through Directional Decomposition University of Maryland, Northeastern University, ByteDance · 2609.15975 · ▲ 9
- The Router Within: Eliciting Native Skill Routing from a Frozen LLM Tsinghua University · 2609.15982 · ▲ 7
- Register Tokens for Bounded-State Reasoning in Diffusion Language Models University of Wisconsin-Madison, Microsoft, University of California San Diego · 2609.16372 · ▲ 7
- RelateAnything: Real-Time Open-Vocabulary Relation Prediction From Any Inputs Independent researchers · 2609.12552 · ▲ 4
- Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration Carnegie Mellon University · 2609.16204 · ▲ 4
- Convergent Emergence of In-Context Learning Across Modalities Johns Hopkins University, Massachusetts Institute of Technology, University of Illinois Urbana-Champaign · 2609.14011 · ▲ 3
- OmniHarness: Harnessing Generalizable Visual Generation via Symbolic Policy Learning Beihang University, Chinese University of Hong Kong, National University of Singapore · 2609.16057 · ▲ 3
- Emergence World: Adversarial Stress-Testing of Long-Horizon Multi-Agent Systems Emergence AI · 2609.17320 · ▲ 3
Tuesday, September 15, 2026 · 33
- Vidu S2: Real-Time Interactive, Editable, and Spatial Video Generation Tsinghua University, Shengshu Technology · 2609.11638 · ▲ 698 · Deep dive
- Atria Dawn: The Dawn of Agentic Superintelligence arXiv · 2609.15818 · ▲ 425
- ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Search Zhongguancun Academy, Zhongguancun Institute of Artificial Intelligence · 2609.13356 · ▲ 257
- Dream-RSI: Recursive Self-Improvement through Evolving Worlds Google, University of Maryland, University of Virginia · 2609.14858 · ▲ 245
- PhysBrain 1.5: From Vision-Language Models to Physical Foundation Models arXiv · 2609.14973 · ▲ 172
- RSIAgent: Autonomous Exploration for Recursive Self-improvement in New Environments Aether AI, University of California San Diego, University of Illinois Chicago · 2609.15364 · ▲ 84
- Grouped Value Attention: Efficient KV Caching via On-Demand Key Reconstruction FrontiersMind · 2609.13285 · ▲ 81
- BVB: Benchmarking Agentic Video Understanding via Programmatic Reconstruction in Blender University of Rochester, Sony, Carnegie Mellon University · 2609.15478 · ▲ 34
- LynnReal-Omni: Native multi-modal Video Generation for Agentic Visual Workflows LynnReal AI · 2609.15863 · ▲ 34
- Discovery Foundation Models: Toward Open-Ended Discovery Intelligence arXiv · 2609.15973 · ▲ 33
- AlayaVista: Streaming World Modeling from Panoramic States to Perspective Video Alaya Lab, Beijing Institute of Technology, University of Tokyo · 2609.14462 · ▲ 26
- How Lossless Is Lossless Speculative Decoding? The Role of Numerical Precision in Orthrus arXiv · 2609.15504 · ▲ 23
- Omni-Streaming Thinking University of Hong Kong, Hong Kong University of Science and Technology, Guangzhou, Lightspeed · 2609.15128 · ▲ 22
- HazardAuditor: From Executable Threats to Safer Computer-Use Agents Ant Group, Zhejiang University, Fudan University · 2609.15134 · ▲ 18
- Kaininja: Extending Native 3D Generators to the Part Level Alaya Lab, University of Tokyo, University of California, Merced · 2609.15659 · ▲ 18
- Agent as Policy for Robotic Manipulation University of Notre Dame, University of California San Diego, San Diego State University · 2609.12541 · ▲ 17
- LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents Inclusion AI, Venus Team, Westlake University · 2609.13287 · ▲ 16
- MInTRL: Off-policy Intervention can boost On-policy RL Amazon, Boston University · 2609.12419 · ▲ 14
- Not All Prompts Are Equal: Exploration-Guided Prompt Scaffolding for Multimodal Reinforcement Post-Training Alibaba, Shanghai Jiao Tong University, Fudan University · 2609.15051 · ▲ 14
- Learning to Solve Hard Problems in RL for LLMs by Never Giving Up Mila, University of Montreal, Allen Institute for AI · 2609.13443 · ▲ 13
- When Agents Slow Down: Understanding LLM Agents' Test-Time Strategies via Elo-per-token Analysis University of California, Berkeley, University of Washington, Princeton University · 2609.15309 · ▲ 10
- Building a Production Greek-English Speech Recognizer Sophea AI Lab at Kiefer · 2609.13498 · ▲ 9
- Enabling Creative Exploration for Vibe Design Agents Google · 2609.15078 · ▲ 9
- Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model Seoul National University · 2609.13053 · ▲ 8
- Pick Your Poison: Learning to Select Poison Sets for Stronger LLM Backdoor Attacks Carnegie Mellon University, Anthropic · 2609.15029 · ▲ 8
- ModaLens: Measuring Image Sensitivity in Report-Conditioned Medical VLMs Massachusetts Institute of Technology, King's College London, American International School Vienna · 2609.15635 · ▲ 8
- Attention-DP3: Spatially Object-aware 3D Diffusion Policy via Geometry-aligned Attentional Conditioning Dalian University of Technology · 2609.13318 · ▲ 7
- Expert-Space Exploration in MoE Reinforcement Learning Tsinghua University, Microsoft · 2609.13058 · ▲ 6
- Root-Cause Attribution Is a Search Problem: Continual Search for Long-Horizon Agent Failures Scale AI · 2609.13463 · ▲ 5
- Thought without systematicity? Evaluating reasoning models on rule induction tasks Princeton University · 2609.13948 · ▲ 3
- E2A-Bench: Benchmarking Evidence-to-Action Reliability in Financial Chart Reasoning Tsinghua University, Jinan University · 2609.14302 · ▲ 3
- Lightning Weave: Improving the Accuracy-Efficiency Frontier of Reasoning Models through Capability Composition Massachusetts Institute of Technology, NVIDIA · 2609.14708 · ▲ 3
- Learning Sparse Decision Trees via Transformer Variational Auto-Encoders University of Pisa, ISTI institute of the Italian National Research Council · 2609.01430 · ▲ 2
Monday, September 14, 2026 · 26
- Benchmark Radar: A Living Database and Search Engine for AI Benchmarks and Evaluation Earth Space AI, Tacite AI, Hangzhou Dianzi University · 2609.11115 · ▲ 171
- DataFlex-RL: An Evaluation Platform for RLVR Data Policies Peking University, University of Chinese Academy of Sciences, Institute for Advanced Algorithms Research, Shanghai · 2609.06107 · ▲ 164
- Feyospace-v1: How the Cyber Mercury Seven Trained Frontier Cyber Models Vera Praxis Lab · 2609.08418 · ▲ 124
- StepAudio 3 Gen Technical Report StepFun · 2609.12945 · ▲ 100
- Occamy-1.0: Open Pareto-frontier 35B Intelligence for Co-work Accio Team · 2609.11977 · ▲ 93
- Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models South China University of Technology, National University of Singapore, Nanyang Technological University · 2609.12641 · ▲ 70
- SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking Tencent, Hong Kong University of Science and Technology, Guangzhou, Hong Kong University of Science and Technology · 2609.13141 · ▲ 65
- ReactHuman: A Physics-Grounded Benchmark for Human-Like Reactive Decision-Making in Embodied Multimodal LLMs University of Montreal, Mila, McGill University · 2609.10895 · ▲ 54
- COBRA-Skills: Contextual Bandit-Guided Evolution for Agent Skill Optimization Chinese University of Hong Kong, Shenzhen, Tianjin University, Hong Kong University of Science and Technology, Guangzhou · 2609.11682 · ▲ 44
- PLC-DPO: Posterior Label Correction in Noisy and Ambiguous Preference Optimization KAIST, KENTECH · 2608.30597 · ▲ 27
- Pelican-Sim 1.0: A General World Model Simulator for Embodied Intelligence Beijing Innovation Center of Humanoid Robotics · 2609.12036 · ▲ 22
- ActionSplice: In-Flight Action Editing for Interactive World Models Texas A and M University · 2609.08230 · ▲ 10
- MobileVLA-R1 2.0: RL-Enhanced Reasoning for Mobile Robot Control Peking University, South China University of Technology, National University of Singapore · 2609.06251 · ▲ 8
- Beyond Top-k Skill Retrieval: Diversity-Aware Skill Routing for LLM Agents Virginia Tech, University of Southern California, Adobe · 2609.05824 · ▲ 7
- Online Learning with LLM Experts from Limited Feedback Virginia Tech, Adobe · 2609.05820 · ▲ 6
- Studying Without a Syllabus: Task-Agnostic Environment Preprocessing Scale AI, University of Maryland · 2609.10824 · ▲ 4
- How Far Can Synthetic Data Take Thai OCR? Wayu Research, Paxa Labs · 2609.03595 · ▲ 3
- Ambient @ EgoProactive 2026 : Proactive Egocentric Assistance with Visually Grounded Supervision Team Ambient · 2609.07099 · ▲ 3
- Ambient @ EgoLongQA 2026: Distilling Long-Video perception into a Sub-2B Model Team Ambient · 2609.07154 · ▲ 3
- TRACE: Trajectory-robust Admission with Evidence Ordering for Efficient GUI Agents Dalian University of Technology, OPPO, Hong Kong Polytechnic University · 2609.10297 · ▲ 3
- SNAP3D: Physically Grounded 3D Parts for Assembly from a Single Image Carnegie Mellon University · 2609.13146 · ▲ 3
- Towards a Deterministic Math Solver for Clinical Language Models MIT Critical Data, University of North Carolina at Chapel Hill, University of Pavia · 2609.10728 · ▲ 2
- Competence-Gated Pooling of Language Models and Priors for Event Forecasting University of Illinois Urbana-Champaign · 2609.12101 · ▲ 2
- Building and Evaluating Fixed-Voice Thai TTS from Synthetic Speech Wayu Research, Paxa Labs, Typhoon · 2609.03502 · ▲ 1
- ReMoMask-2: Latent Retrieval-Augmented Masked Motion Generation University of Sydney, Peking University, University of Chinese Academy of Sciences · 2609.08365 · ▲ 1
- Feature Recovery for Object Understanding After Irreversible Fire Damage University of Illinois Urbana-Champaign · 2609.12078 · ▲ 1
Friday, September 11, 2026 · 26
- NCP-ArchPreview Technical Report: Moving towards Latent Space Language Models through Next Concept Prediction Shanghai AI Laboratory, Shanghai Jiao Tong University · 2609.10715 · ▲ 328 · Deep dive
- SenseNova-U1.5: Towards Native Unified Visual Intelligence arXiv · 2609.11929 · ▲ 273
- SpatialBlock: Enhancing Spatial Intelligence in LVLMs via Synthetic Block-Stacking Problem KAIST, AITRICS · 2609.07064 · ▲ 143
- Mi-Ripple: Restoring Images Degraded by Iterative AI Editing Miyang Technology, Key Laboratory of System Software of the Chinese Academy of Sciences, Institute of Software of the Chinese Academy of Sciences · 2609.11317 · ▲ 52
- X-AuT: Progressive Audio-Encoder Compression for Speech LLMs with Cross-Scale Distillation XPeng · 2609.11412 · ▲ 46
- An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics NVIDIA · 2609.10712 · ▲ 43
- EvoSafeHarness: Evolving Model- and Domain-Specific Harnesses for Securing Agents Johns Hopkins University, University of Wisconsin-Madison, NVIDIA · 2609.05903 · ▲ 40
- Memory as Plans: World-Action Modeling with Memory-Grounded Planning Harbin Institute of Technology, Nanyang Technological University, Shandong University · 2609.11561 · ▲ 40
- FreeFlow: A Bias-free Hierarchical Transformer for Optical Flow Estimation Lomonosov Moscow State University, Moscow State University Institute for Artificial Intelligence · 2609.11486 · ▲ 34
- Beyond Solver Verdicts: Generative Reward Models for Autoformalization Case Western Reserve University, Amazon · 2609.11085 · ▲ 33
- Recursive Code World Models: Building Complex Worlds through Recursive Scene Programs Georgia Institute of Technology · 2609.11499 · ▲ 33
- Negative Self-Distillation: Learning to Reason by Avoiding Flaws University of Virginia, Stanford University · 2609.11699 · ▲ 33
- Building Multilingual Bridges: Data Mixing as the Pillar of Generalization for In-Language Reasoning Cohere Labs, Cohere · 2609.10445 · ▲ 32
- World in World: Explore the World with World Models Westlake University · 2609.11548 · ▲ 31
- TempCloze: Can Video-LLMs Identify the Missing Middle? University of Hong Kong, National University of Singapore, Peking University · 2609.01515 · ▲ 30
- HyQuant: Hybrid-Precision Quantization for LLM Attention Shanghai Jiao Tong University, Xi'an Jiaotong University, Tencent · 2608.27875 · ▲ 28
- Studying Image Tokenizers as Visual Languages in Unified Multimodal Models University of Washington, Amazon · 2609.09143 · ▲ 28
- MetroLLM-Bench: Evaluating Language Models as Transit Kiosk Runtimes Continker · 2609.10016 · ▲ 28
- DRG-MAPPO: Hierarchical Dynamic Role-Graph Multi-Agent Reinforcement Learning for Cooperative Air Combat Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences · 2609.11155 · ▲ 28
- Generative Late-Interaction Embeddings For Visual Document Retrieval King Abdullah University of Science and Technology, Edge Hill University · 2609.11808 · ▲ 26
- Adaptive Bridge: A Proxy-Based Decoupling Layer for Mitigating DDS Backpressure in ROS 2 International Institute of Information Technology Bangalore · 2608.15380 · ▲ 25
- CARDEA: Auditable Reasoning Grounded in Spatial Evidence for End-to-End Coronary Angiography Interpretation China Medical University Hospital, China Medical University · 2609.06931 · ▲ 25
- IdeaAMBIG: Benchmarking Implementation-Critical Gaps in Research-Idea Specifications Yale University, TCS Research · 2609.10539 · ▲ 24
- ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation Yale University, University of Chicago, TCS Research · 2609.09076 · ▲ 23
- UniH^3: Unifying Hierarchical Homogeneity and Heterogeneity for All-in-One Medical Image Restoration Beihang University, Tsinghua University, Independent researchers · 2609.11156 · ▲ 23
- Think Before You Link: Rarity, Reasoning, and Retrieval in Multilingual Entity Linking Carnegie Mellon University · 2609.10745 · ▲ 20
Thursday, September 10, 2026 · 32
- Scaling Automatic Research Agents via World Models University of Illinois Urbana-Champaign, Amazon · 2608.12564 · ▲ 477 · Deep dive
- Show-Harness: Just a VLM Agent Can Play Robots National University of Singapore · 2609.10522 · ▲ 160
- Programmable World Model Alaya Lab · 2609.10540 · ▲ 113
- AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems Korea University, KAIST, Meta · 2609.08572 · ▲ 82
- T1: Terminal Agent Reinforcement Learning for Long-Horizon Tasks Tencent, National University of Singapore, University of Georgia · 2609.11042 · ▲ 62
- WearableQA: A Benchmark for Health Reasoning over Real-World Wearable Data Meta, KAIST, Korea University · 2609.05405 · ▲ 30
- PARSER: Read in Parallel, Reason in Depth for Long-Context LLM Agents Chinese University of Hong Kong · 2609.06702 · ▲ 25
- Revisiting Complete Reasoning Traces for Post-Training NAVER AI Lab · 2609.07103 · ▲ 22
- Puppeteer: Object-Grounded Posture-Aware Co-Speech Gesture Generation Pickford AI, University of Toronto, Vector Institute · 2609.00369 · ▲ 20
- SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research? Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences · 2609.09113 · ▲ 19
- Φ-Bench: Can Large Language Models Engineer the Infrastructure That Powers Them? University of Science and Technology of China, StepFun, Peking University · 2609.10226 · ▲ 18
- Why Is Video Still So Expensive? A Survey of Inference-Efficiency Mechanisms in Video and Audiovisual LLMs Télécom SudParis, Institut Polytechnique de Paris, Moments Lab · 2609.10355 · ▲ 17
- SyncWorld: Visual Calibration Enables World Models as Zero-Shot Simulators University of Massachusetts Amherst, University of California, Berkeley, New York University · 2609.09155 · ▲ 15
- Scores Alone Do Not Prove Discovery: The Discovery Certification Protocol for Auditing AI Research Agents Carnegie Mellon University · 2609.09219 · ▲ 15
- SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering Agents East China Normal University, Shanghai AI Laboratory, Fudan University · 2609.08149 · ▲ 13
- AgenticGen: Reward-Guided Agentic Video Generation for Advertising ByteDance, Tsinghua University, Chinese University of Hong Kong · 2609.09187 · ▲ 13
- DianShi-RxnDB: A Large-Scale, Fine-Grained Organic Reaction Data Platform Built via a Fully Automated Pipeline for Researchers and AI Agents Shanghai AI Laboratory, Fudan University, Shanghai Jiao Tong University · 2609.06703 · ▲ 12
- Co-Evolving Harnesses and Models: On-Policy Correction Helps Weaker Models Catch Up Where Imitation Fails Salesforce · 2609.09134 · ▲ 11
- SchemeArena: Factorized Stress Testing of Scheming in LLM Agents University of Michigan · 2609.08126 · ▲ 10
- Difficulty-Adaptive Tree-Structured Policy Optimization for Expanding Reasoning Coverage in RLVR POSTECH · 2609.08650 · ▲ 10
- PlannerForge: LLM Agents for Scenario-Based Testing of Motion Planners in Autonomous Driving Technical University of Munich, University College London · 2609.08965 · ▲ 8
- StochBench: A Domain-Specific Benchmark for Stochastic Processes in Lean Case Western Reserve University · 2609.09264 · ▲ 8
- Diffs vs. Whole Files: An Empirical Comparison of Iterative Edit-Based and Direct Generation for Flutter/Dart Code Models Independent researchers · 2609.05779 · ▲ 7
- OracleZoom: On-Policy Self-Distillation Inspired Reference-Constrained Recursive Image Super Resolution University of Maryland, Baltimore County · 2609.06490 · ▲ 7
- A Three-Layer Caching Architecture for Low-Latency LLM Web Search on Commodity CPU Hardware pollinations.ai · 2609.05463 · ▲ 6
- Train Smarter, Not Harder: Switching Signal-Guided Training in Active Learning Technion · 2609.06806 · ▲ 5
- The Semantic Bottleneck: Leveraging Semantic Representations for Non-Invasive Speech Decoding University of Oxford · 2609.10296 · ▲ 5
- The Price of Sparsity: Sufficient Conditions for Sparse Recovery using Sparse and Sparsified Measurements Massachusetts Institute of Technology · 2509.01809 · ▲ 4
- DF26: We Cannot Tell Fake From Real Anymore Ukrainian Catholic University, Hover, Czech Technical University in Prague · 2609.07369 · ▲ 4
- From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data Attribution Peking University, YiXin AI Lab, Beijing Academy of Artificial Intelligence · 2609.02771 · ▲ 3
- RESCUE-BENCH: Towards Relation-Aware Multi-Party Emotional Support Conversation Systems Hong Kong Polytechnic University, Vrije Universiteit Amsterdam, Nanjing University of Science and Technology · 2609.09657 · ▲ 3
- Reference-Based Bias Detection in LLMs via Relative Representations of Hidden States NASK National Research Institute, Warsaw University of Technology, Gdansk University of Technology · 2609.10060 · ▲ 3
Wednesday, September 9, 2026 · 48
- NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness arXiv · 2609.08183 · ▲ 173
- AuK Technical Report: An Open-Source Foundational Model for Speech Generation and Editing arXiv · 2609.08936 · ▲ 164
- Omni Interaction Agent Technical Report Tencent, Zhejiang University, Shanghai Jiao Tong University · 2609.08977 · ▲ 95
- Eliciting Weak-to-Strong Generalization with On-Policy Reverse Distillation KAIST, Microsoft, University of Toronto · 2609.08798 · ▲ 81
- OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining National University of Singapore, Tsinghua University, Peking University · 2609.07398 · ▲ 76
- Marigold V2: Revisiting Diffusion Transformers for Monocular Depth Estimation EPFL, Huawei, University of Bologna · 2609.08084 · ▲ 70
- Miles v0.1: Production-Level Post-Training RadixArk · 2609.08368 · ▲ 58
- DriveZero: End-to-End Driving Beyond Human Demonstrations Xiaomi · 2609.06055 · ▲ 56 · Deep dive
- GE-Act 2.0: Pretraining and Scaling a World-Action Model for Robotic Manipulation AgiBot · 2609.05588 · ▲ 56
- Mask Forcing: Improving Autoregressive Video Diffusion Distillation via Dual-Noise Masking Rollout Hong Kong University of Science and Technology, Guangzhou, Hong Kong University of Science and Technology, Lightspeed · 2609.09123 · ▲ 54
- BeaconKV: Key-Value Cache Compression Guided by Beacon Queries for Efficient Large Reasoning Model Inference Hanyang University, Sungkyunkwan University · 2609.04971 · ▲ 42
- Procedural Graphs: Self-Evolving Execution Structures for LLM Agents Google, Georgia Institute of Technology, Peking University · 2609.09153 · ▲ 41
- Reason Through the Latent! Making Latent Visual Reasoning Necessary Korea University, AIGEN Sciences · 2609.06746 · ▲ 36
- SceneMosaic: Efficient and Diverse Simulation-Ready Scene Generation via Hybrid Agentic Layout Evolution University of Hong Kong, University of Electronic Science and Technology of China · 2609.05594 · ▲ 35
- Online Draft Co-Training for Speculative Decoding in Large-Scale, Long-Context RL Post-Training NVIDIA · 2609.07108 · ▲ 35
- Agentic Visual Generation: From Generative Models to Agentic Control Fudan University, Shanghai Innovative Institute, Chinese University of Hong Kong MMLab · 2609.06758 · ▲ 33
- CosmoH2G: A Hand-to-Gripper Transfer Dataset and Baseline Method for Object Manipulation with Complex Spatial Movements Chinese University of Hong Kong, Shenzhen, GenuX, Future Network of Intelligence Institute, Shenzhen · 2609.07498 · ▲ 33
- EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness? Salesforce, University of North Carolina at Chapel Hill, University of Wisconsin, Madison · 2609.04280 · ▲ 30
- Kalman Delta Networks: Uncertainty-aware Associative Memory Yale University · 2609.07816 · ▲ 29
- RoboSPA: Can VLA Models Go Beyond Simple Scenes and Short-Horizon Tasks? Zhejiang University, University of Electronic Science and Technology of China, South China Normal University · 2609.05324 · ▲ 28
- VDiff-Bench: A Challenging Benchmark for Fine-Grained Image Difference Identification University of California, Los Angeles · 2609.06245 · ▲ 27
- Steering Geometry: Validating Human Value Geometry in LLM Steering Space University of British Columbia, Vector Institute, Queen's University · 2609.06289 · ▲ 27
- What Did I Just Say? Self-Listening for Full-Duplex Speech Models Shenzhen Loop Area Institute, Chinese University of Hong Kong, Shenzhen · 2609.05592 · ▲ 23
- CoVeR: Coverage-Based Token Pruning for Multi-View 3D Reasoning in VLMs Carnegie Mellon University, Meta · 2609.08345 · ▲ 23
- Environments as Scaffold: Enriching Feedback to Bootstrap Self-Evolving Agents in Long-Horizon Tasks Fudan University, Institute of Automation, Chinese Academy of Sciences, Shanghai Innovation Institute · 2609.08404 · ▲ 23
- Measuring Language Transfer in Robot Policies: Adding Greek to a Cosmos3 Vision-Language-Action Policy Sophea AI, Kiefer SA · 2609.07470 · ▲ 22
- TANGO: Humanoid Navigation in Cluttered Environments with a Whole-Body Vision-Language-Action Model University of California, Berkeley, Peking University, Tsinghua University · 2609.09158 · ▲ 21
- What LLM Trading Agents Actually Do in Production: A Six-Month, Population-Scale Record from Two Fleets DX Research Group · 2609.05663 · ▲ 20
- VidaForge: Open Research Infrastructure for Video Pretraining Data Recipes Shanghai Jiao Tong University, Shanghai Innovation Institute, Fudan University · 2609.06652 · ▲ 19
- TransNormal-2: Geometry-Grounded Rectified Flow with Edge-Aware Decoding for Precise Normal Estimation Zhejiang University, Zhongguancun Academy · 2609.06665 · ▲ 19
- SQS: Bayesian DNN Compression through Sparse Quantized Sub-distributions Purdue University, University of Texas at El Paso · 2510.08999 · ▲ 18
- Cadence: Error-Bounded Lossy Compression of Demand Time Series with a Time-Series Foundation Model Independent researchers · 2609.06008 · ▲ 18
- Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable Questions University of Southern California, Arizona State University · 2608.29109 · ▲ 17
- MOLE: Detecting Insider Threats in AI Agents Carnegie Mellon University · 2609.06966 · ▲ 17
- Harnessing CLIP and DINO: An Uncertainty-Aware Cascaded Fusion Network for Generalizable Deepfake Image Detection Beijing Jiaotong University, Tsinghua University, Ant Group · 2609.07670 · ▲ 17
- ReactVAU: A Slow-Fast Decoupled Framework for Streaming Video Anomaly Understanding National Tsing Hua University, NVIDIA · 2609.07941 · ▲ 17
- SynthGait-19K: A Physically Grounded Synthetic Video Dataset for Gait Parameter Estimation KITE Research Institute, University of Toronto, Vector Institute · 2609.08108 · ▲ 17
- StudyBench: Can Self-Evolution Squeeze Textbooks for Olympiad Capability? Tsinghua University, Zhejiang University · 2609.00787 · ▲ 16
- Multi-Grid Post-Training for Long-Form Multi-Shot Video Generation University of California, Santa Cruz, University of Florida, Vanderbilt University · 2609.06373 · ▲ 16
- Encoded Early, Used Late: Where Transformers Begin to Act on an Inferred Partner's Expertise Georgia Institute of Technology, Northeastern University · 2609.07139 · ▲ 15
- A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLM Beijing University of Posts and Telecommunications, Hong Kong Polytechnic University, Tsinghua University · 2609.07821 · ▲ 15
- RelightFormer: Feed-forward Generative Transformer for Multiview Object Relighting Hong Kong Polytechnic University · 2609.07414 · ▲ 10
- Graph Machine: Towards Better Pretraining via Edges arXiv · 2609.02881 · ▲ 6
- RenderFormer-V2: Neural Rendering with Heterogeneous Scene Primitives Stanford University, Microsoft, College of William and Mary · 2609.05738 · ▲ 5
- MasterControl Seventeen Every Time MasterControl AI Lab · 2609.03209 · ▲ 3
- Counter-Swarm Doctrine: Containing Coordinated Agent Intrusions MoltAI Corp · 2609.06140 · ▲ 3
- NOAH: Learning the Full Patient Journey. A Longitudinal Multimodal Time-Aware Model for Representation and Forecasting TUM University Hospital, Technical University of Munich, University Medical Center Hamburg-Eppendorf · 2609.09140 · ▲ 3
- Learning 3D Editing without Paired Supervision via Generative Prior Distillation Beihang University, Central University of Finance and Economics, VAST · 2609.04942 · ▲ 1
Tuesday, September 8, 2026 · 12
- Unlocking Lossless Speedups in LLMs via Discrete Diffusion Institute of Foundation Models, University of Illinois Urbana-Champaign, Cornell Tech · 2609.04010 · ▲ 105
- FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience Tencent, University of Pennsylvania · 2609.03241 · ▲ 53
- Causal Foundation Models Layer 6 AI, TD Bank Group · 2609.03003 · ▲ 28
- ENEAS: Embedding-guided Neural Ensemble for Adaptive Segmentation SperidLabs · 2609.03756 · ▲ 24
- Verify Before You Distill: Prompt-Level Teacher Gating for On-Policy Distillation AllSpark Team · 2609.02998 · ▲ 16
- One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation OVHAILLM · 2608.25936 · ▲ 15
- EmbodiedSkills: A Unified Framework for Orchestrating, Training, and Deploying VLA Agents Zhejiang University, Nanjing University of Aeronautics and Astronautics, Cornell University · 2609.01281 · ▲ 14
- SimpleMemVLA: A Simple but Effective Native-Video Memory for Vision-Language-Action Models Huazhong University of Science and Technology, Renmin University of China, Tsinghua University · 2609.05533 · ▲ 14
- Unifying Conformal Language Tasks with In-Context Ensembles Signal 1 AI, Layer 6 AI · 2609.03005 · ▲ 10
- Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal Multiverse Computing · 2609.04482 · ▲ 10
- What Else Needs Fixing? Exploring Cost-Effective Test-Time Compute for Revision Propagation in Artifacts Generated Through Conversation NTT · 2609.03254 · ▲ 6
- Privacy Failure in Split-LLM Training, The Returned Gradient Nullifies the Decoys Setloop.io · 2609.04382 · ▲ 3
Monday, September 7, 2026 · 26
- Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems UCL Centre for Artificial Intelligence, University of Liverpool, Huawei · 2609.02750 · ▲ 91 · Deep dive
- Iris: Climbing to the Search Frontier arXiv · 2609.04304 · ▲ 50 · Deep dive
- Motion-Omni: End-to-End Joint Speech and Full-Body Motion for Spoken Dialogue Peking University, Lightspeed, Chinese University of Hong Kong, Shenzhen · 2609.04250 · ▲ 35 · Deep dive
- The Attention Triangle in Audio-Video Models Tel Aviv University, Simon Fraser University · 2609.03586 · ▲ 25 · Deep dive
- WorldSculpt: Generating Compositional Worlds from Grounded Videos Alaya Lab, University of Tokyo · 2609.05416 · ▲ 21 · Deep dive
- Enoki: Efficient Multi-Level Hallucination Detection Skoltech, Artificial Intelligence Research Institute, Sber AI Lab · 2609.00581 · ▲ 16 · Deep dive
- Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization Cardinal Operations, Shanghai Jiao Tong University · 2609.05258 · ▲ 15 · Deep dive
- Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference Cerebras Systems, Mohamed bin Zayed University of Artificial Intelligence · 2609.05275 · ▲ 13 · Deep dive
- Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs KAIST, NAVER AI Lab · 2609.04753 · ▲ 10 · Deep dive
- MaxKernel: Agentic Kernel Generation for TPUs Google · 2609.04523 · ▲ 9 · Deep dive
- RISE: Recursive Improvement via Self-Extrapolating Policy Distillation Salesforce · 2609.05295 · ▲ 9 · Deep dive
- UniMate: One Unified Model to Animate Diverse Skeletons Princeton University, University of California, Berkeley, Massachusetts Institute of Technology · 2609.05415 · ▲ 9 · Deep dive
- One Editor, Many Edits: A Unified Training-Free Framework for Diverse Video Editing University of Illinois Urbana-Champaign · 2609.04190 · ▲ 6 · Deep dive
- Dr. Claw: An AI Scientist Workspace for Vibe Research Lehigh University, University of Illinois Chicago, University of Pennsylvania · 2609.00365 · ▲ 5 · Deep dive
- Group Adaptive Clipping Policy Optimization University of Toronto, Amazon · 2609.00444 · ▲ 5 · Deep dive
- When Models Edit Too Much: On the Fidelity of Minimal Code Edits National University of Singapore · 2609.04061 · ▲ 4 · Deep dive
- Training-Free Speech-Centric Omni Understanding with Frozen VLMs Mohamed bin Zayed University of Artificial Intelligence · 2609.04242 · ▲ 4 · Deep dive
- τ^τ-Bench: An Environment for End-To-End, Realistic Agent Construction Sierra, Princeton University · 2609.04611 · ▲ 4 · Deep dive
- To See a World in a Living Context: Unified Indoor-Outdoor Urban World Generation Sun Yat-sen University, Norwegian University of Science and Technology · 2608.05879 · ▲ 3 · Deep dive
- ShallowStream: Index Shallow then Answer Deep for Streaming Video Understanding Harbin Institute of Technology, Shenzhen · 2609.02780 · ▲ 3 · Deep dive
- When Quantization Breaks Memory: Recurrent-State Write-Back in Low-Precision Temporal Inference Rensselaer Polytechnic Institute · 2609.04490 · ▲ 3 · Deep dive
- AdaptVPR: Route-Aware Hard Positive Generation for Robust Visual Place Recognition Beijing University of Posts and Telecommunications, Qilu University of Technology · 2609.04369 · ▲ 2 · Deep dive
- The 2026 PNPL Competition: Word Classification and Efficient Cross-Subject Generalisation in LibriBrain100 University of Oxford, Maastricht University, University of Montreal · 2609.03231 · ▲ 1 · Deep dive
- HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals Compassion Aligned Machine Learning, University of Warwick · 2609.04444 · ▲ 1 · Deep dive
- Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Models POSTECH · 2609.04714 · ▲ 1 · Deep dive
- Knowing What Not to Answer: Selective Non-Compliance in Vision-Language Models POSTECH · 2609.04720 · ▲ 1 · Deep dive
Friday, September 4, 2026 · 27
- Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments Alibaba, Tsinghua University · 2609.04148 · ▲ 150 · Deep dive
- Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training Alibaba · 2608.26730 · ▲ 145 · Deep dive
- LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes AGI Research Center at Inclusion AI · 2609.03796 · ▲ 111 · Deep dive
- Compile by Training: Turning Natural-Language Specifications into Local Neural Functions University of Waterloo, Harvard University · 2609.04199 · ▲ 84 · Deep dive
- LatentPress: Context Compression Beyond Text and Vision Cornell University, Arizona State University · 2609.01507 · ▲ 74 · Deep dive
- Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning Salesforce, University of Illinois Urbana-Champaign · 2609.03430 · ▲ 71 · Deep dive
- Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM arXiv · 2609.04098 · ▲ 65 · Deep dive
- Rethinking On-Policy Distillation of Large Language Models II: One Training Example Tsinghua University, University of Chinese Academy of Sciences, Northeastern University · 2609.04172 · ▲ 61 · Deep dive
- Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States Nanyang Technological University, University of Michigan, Beijing Jiaotong University · 2609.04196 · ▲ 54 · Deep dive
- Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction National Yang Ming Chiao Tung University, NVIDIA · 2609.04201 · ▲ 30 · Deep dive
- Editable Visual Design Tencent, Sun Yat-sen University, Tsinghua University · 2609.04034 · ▲ 30 · Deep dive
- The Missing Temporal Link: Temporal Context Routing for Script-Driven Audio-Video Generation Peking University, Alibaba, Hong Kong University of Science and Technology · 2609.02367 · ▲ 25 · Deep dive
- Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding Nanjing University of Science and Technology, Ant Group, National University of Singapore · 2609.04131 · ▲ 22 · Deep dive
- CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation Institute of Automation at the Chinese Academy of Sciences, Alibaba, University of Chinese Academy of Sciences · 2609.04083 · ▲ 17 · Deep dive
- RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests Sungkyunkwan University · 2608.27831 · ▲ 16 · Deep dive
- WorldReward: Reward Modeling for Camera-Conditioned World Models Fudan University, Tencent, Shanghai Innovation Institute · 2609.03952 · ▲ 15 · Deep dive
- PACE: Towards Surfacing Hidden Conflicts in User Requests POSTECH · 2609.03293 · ▲ 15 · Deep dive
- Last Translation Benchmark ETH Zurich, Johns Hopkins University, Mohamed bin Zayed University of Artificial Intelligence · 2609.04173 · ▲ 12 · Deep dive
- FlashRender: Few-Step Generative Rendering via Camera-Controlled Video MeanFlow EverEx, Yonsei University, Korea University · 2609.03563 · ▲ 11 · Deep dive
- Environment Evolution for Terminal Agents Tencent · 2609.04128 · ▲ 10 · Deep dive
- Using Grounded Theory for Agent Behavior Analysis at Scale Purdue University, Stevens Institute of Technology, Cornell University · 2608.30391 · ▲ 10 · Deep dive
- Principia: Relational Physics Tests for Video Models Indian Institute of Science, Johns Hopkins University · 2609.04200 · ▲ 8 · Deep dive
- Select, Compress, Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLMs Independent researchers · 2609.03820 · ▲ 8 · Deep dive
- Percolation Dynamics in Optimization : Variance Cascades and Discrete Scale Invariance Technical University of Munich · 2609.02373 · ▲ 7 · Deep dive
- Let Confidence Change, Not the Prediction: Prediction-Preserving Repair for Post-hoc Calibration Hanyang University, Hankuk University of Foreign Studies · 2609.01072 · ▲ 7 · Deep dive
- QCell: Recombining and Aligning Cell Queries for Overlapping Instance Segmentation University of Tartu, Ukrainian Catholic University · 2608.29253 · ▲ 7 · Deep dive
- DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training Carnegie Mellon University, IBM · 2609.04094 · ▲ 1 · Deep dive
Thursday, September 3, 2026 · 36
- Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills Beijing Academy of Artificial Intelligence, University of Science and Technology of China, Renmin University of China · 2609.02749 · ▲ 512
- HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness? ByteDance, Singapore University of Technology and Design, Georgia Institute of Technology · 2609.01437 · ▲ 236
- Aspire: Can Models Self-Evolve from Vague Goals? ByteDance, Singapore University of Technology and Design, M-A-P · 2608.31111 · ▲ 210
- SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models Chinese University of Hong Kong, Shenzhen, SLAI, National University of Singapore · 2609.02886 · ▲ 135 · Deep dive
- EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction Shanghai Jiao Tong University, Singapore Management University, East China Normal University · 2609.02783 · ▲ 113
- It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning University of North Carolina at Chapel Hill, Apple · 2609.00638 · ▲ 68
- Language Models Can Control Their Own Attention KAIST, Google · 2609.02737 · ▲ 59
- On the Design Fundamentals of Pixel Text Representation Learning Chinese University of Hong Kong, Alibaba, Fudan University · 2609.01147 · ▲ 32
- S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement? ByteDance, M-A-P, TokenWave AI · 2608.31100 · ▲ 29
- WHALE: A Simple Recipe for Joint Harness-Weight Optimization Krafton, KAIST, Stanford University · 2609.00196 · ▲ 29
- Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering Tsinghua University, University of Arizona · 2608.21450 · ▲ 26
- NeoMME: A Single-Tower Multimodal-Native Multilingual Foundation Encoder for Efficient Fine-Tuning and Inference H Company · 2609.01657 · ▲ 24
- ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes Zhejiang University, Monash University, University of Adelaide · 2609.01740 · ▲ 23
- A Glance Is All You Need: Single-Pass Fine-Grained Image Captioning with SimLoss University of Massachusetts Amherst, Adobe · 2609.00591 · ▲ 16
- Cliff: Learning Process Rewards from the First Mistake Amazon, University of Illinois Urbana-Champaign · 2609.02817 · ▲ 15 · Deep dive
- Influence-Directed Distillation: Solving the Diversity Bottleneck in Sampled-Token On-Policy Distillation Bilibili, University of North Carolina at Chapel Hill, University of Science and Technology of China · 2608.29846 · ▲ 15
- VibeVoice-ASR-Streaming Technical Report Microsoft, University of Chinese Academy of Sciences, Shanghai Jiao Tong University · 2609.02812 · ▲ 13
- Post-Training Language Models for Gold-Medal Performance in Coding Competitions NVIDIA · 2609.02849 · ▲ 9 · Deep dive
- Institutional Newspapers Pipeline: Deriving billions of high quality tokens from historical newspapers Institutional Data Initiative at Harvard Law School Library, Boston Public Library, Harvard Law School Library · 2608.18972 · ▲ 9
- MULTI3IR: A Benchmark for Multi-perspective Multi-domain Multi-modal Information Retrieval Seoul National University · 2608.30949 · ▲ 8
- CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing University of Oregon, Adobe, Hanoi University of Science and Technology · 2609.01925 · ▲ 8
- ExecRetrieval: Measuring the Functional-Correctness Gap in Code-Embedding Retrieval Kennesaw State University · 2609.01865 · ▲ 6
- PaperCompiler: Faithful Paper-to-Code Generation via Repository-Level Specification Compilation Nanyang Technological University, Singapore · 2609.02272 · ▲ 6
- SnapBench: Benchmarking Snap-and-Ask Multimodal Retrieval for Mobile Interactions Hong Kong University of Science and Technology, Guangzhou, Tencent, Tsinghua University · 2608.29607 · ▲ 5
- Exploring Collaboration between a language and a non-language agent Adobe, Indraprastha Institute of Information Technology Delhi, Indian Institute of Technology Kanpur · 2609.00474 · ▲ 5
- Ignorance or Incompetence? Constructing Knowledge-Gated, Verifiable Tasks for LLM Agents DataGrids, Shanghai University, Peking University · 2608.30322 · ▲ 4
- Autoregressive Mosaics: Probing 2D Spatial Reasoning in Text-Only Language Models University of Rostock · 2608.30751 · ▲ 3
- FoldingAgent: Inferring Parametric Origami Procedures from Demonstration Videos Weizmann Institute of Science, Massachusetts Institute of Technology · 2609.00377 · ▲ 3
- Kirin: Animal Motion Generation from In-the-Wild Video University of Illinois Urbana-Champaign, University of Pennsylvania, Stanford University · 2609.01823 · ▲ 3
- Replacing Training with Memory: Listwise Selection for Text-to-SQL Seoul National University, KAIST · 2609.00834 · ▲ 2
- Sparse Readout Prism: Explaining Logit-Lens Scores in Features Instead of Tokens University of Cambridge · 2609.01936 · ▲ 2
- Debias-SparseGPT: Bias-Aware Pruning for Large Language Models Hubert Curien Laboratory, Claude Bernard University Lyon 1, Lumière University Lyon 2 · 2609.02496 · ▲ 2
- An Empirical Study on Zero-Data Bootstrapping for Conversational Recommender Systems University of California San Diego, Netflix, Cornell University · 2504.15476 · ▲ 1
- Portfolio Risk Bounds without Cross-Asset Return Covariances: Distributional Fields from Language-Model Representations University of Cape Town · 2608.29692 · ▲ 1
- Small Language Models as Judges for Rubric-Based Reinforcement Learning New York University, Yale University · 2608.30005 · ▲ 1
- Wasserstein-Barycentric Interaction Fields for Spatial Factor Models: Evidence from Language-Model Representations University of Cape Town · 2608.29669
Wednesday, September 2, 2026 · 29
- StudentSim: Training LLM-based Student Simulators Microsoft, University of Illinois Urbana-Champaign · 2609.01591 · ▲ 480
- Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving Alibaba, Huazhong University of Science and Technology · 2609.00111 · ▲ 368
- SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers Tsinghua University, ByteDance, M-A-P · 2609.01343 · ▲ 90
- UI-Venus-2 Technical Report Ant Group · 2609.00028 · ▲ 59
- H3-World: Turning Language Understanding into World Control Tencent, National University of Singapore, Hong Kong Polytechnic University · 2609.01560 · ▲ 47
- ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training Joy Future Academy · 2609.00188 · ▲ 45
- From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix T-Tech · 2609.01572 · ▲ 33
- Hi-Q: Hierarchical Evidence-guided Query Refinement for Multi-Hop Question Answering POSTECH · 2608.30468 · ▲ 32
- Evaluating Multimodal LLMs as Generalist Vision-Language-Action Agents for Drone Control: Commanding, Approaching, Tracking and Searching Naver Cloud · 2609.01404 · ▲ 26
- Uncovering Understanding-Generation Synergy in Native Unified Multimodal Models: From Representation, Task to System Nanyang Technological University, SenseTime · 2609.01607 · ▲ 23
- Safin-1: Safety from Within through Memory-Native State Evolution Shanghai AI Laboratory · 2609.00092 · ▲ 20
- AgentJudgeBench: A Multi-Difficulty Benchmark for Evaluating LLM Judges on Agentic Tool-Calling ServiceNow AI · 2608.26623 · ▲ 19
- DiagEvo: Diagnosis-Guided Self-Evolution via Hierarchical Error Memory Chinese University of Hong Kong, Shenzhen, LongCat Team at Meituan, Peking University · 2609.00768 · ▲ 18
- Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement Shanghai AI Laboratory · 2609.01481 · ▲ 16
- E-Commerce Bench: Evaluating LLM Agents on Long-Horizon Autonomous Business Operation Alibaba, Hong Kong University of Science and Technology · 2608.30730 · ▲ 15
- EM^2Mem: Event-Centric Multimodal Memory for Large Language Models Zhejiang University, South China University of Technology, Lenovo · 2609.00551 · ▲ 13
- Agents in the Large: Perception-Centered Architecture for Persistent Agents Fudan University, Peking University, Nanyang Technological University · 2608.30478 · ▲ 10
- InternReviewer & InternAdvocate: Objective Reward and Evaluation for Agentic Reinforcement Learning in Peer Review and Rebuttal Shanghai AI Laboratory, Beijing Jiaotong University, Tsinghua University · 2608.28612 · ▲ 8
- Learning Where Outcomes Change:Credit-Addressable Reasoning for Multimodal Geometry Tsinghua University, Microsoft, Zhejiang University · 2608.30457 · ▲ 8
- Control-Data Flow Separation: Stable Prompt Optimization in Multi-Agent LLMs University of Waterloo, Manulife · 2609.00621 · ▲ 8
- Recursive Criticality of AI Self-Improvement London Institute for Mathematical Sciences · 2609.00137 · ▲ 6
- Adapting Without Gradients: Affine Statistics Transport and What Its Certificate Can Tell You Talan Research Center · 2609.00374 · ▲ 6
- Agent Memory Is a Surface for Endogenous Authorization Laundering ETH Zurich, Georgia Institute of Technology · 2609.01836 · ▲ 6
- The Mechanics of Democratic Dominance: A System Dynamics Paradigm for Dynamic Consent Engineering Asia School of Business · 2608.27509 · ▲ 5
- DramaChain Bench: An End-to-End Benchmark for Short-Drama Generation Tencent, Beijing Film Academy, Peking University · 2609.00646 · ▲ 5
- ReFlowSET: Representation-Aligned Latent Flow Matching for SAR-to-EO Image Translation KAIST, Stellarvision · 2609.00968 · ▲ 5
- Token-Efficient Data Reasoning Agents via Adaptive Structuring of Unstructured Data Harvard University · 2608.31082 · ▲ 4
- Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall Meta, University of Washington, Princeton University · 2609.01532 · ▲ 4
- Does Imitation Learning Preserve Temporal Robustness in Dexterous Manipulation? An Expert-Learner Comparison Across Task Execution Speeds University of Maryland · 2609.01453 · ▲ 2
Undated · 4
- When Fine-Grained Credit Assignment Helps... University of Pennsylvania, Philo Labs · 2610.35497
- DeCoPrune: Efficient KV-Cache Pruning for... Nanyang Technological University, Tsinghua University, Princeton University · 2609.39096
- ControlScope: Workflow Revision and... Carnegie Mellon University, Tsinghua University · 2609.34313
- CineSubBench: Evaluating LLMs on Long-Form... University of Alberta, Independent researchers · 2609.36218