Daily Papers of 2026-10-21

  1. Focused Omni-Modal GUI Agents 230 upvotes, #1 of 2026-10-21
  2. Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds 197 upvotes, #2 of 2026-10-21
  3. PAR: Progress-Aligned Recurrence for Length Generalization in State Tracking 167 upvotes, #3 of 2026-10-21
  4. Ropedia W0: Scaling Loco-Manipulation World Models across Embodiments with Human Experience 151 upvotes, #4 of 2026-10-21
  5. CoEvo-RSI: When Harnesses Evolve, Which Models Should Follow? 120 upvotes, #5 of 2026-10-21
  6. S-Agent 2.0: Spatial Reasoning with a General Coding Agent 109 upvotes, #6 of 2026-10-21
  7. DecisionPivot: A Diagnostic Benchmark for Decision Models in LLM Systems 74 upvotes, #7 of 2026-10-21
  8. Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition 74 upvotes, #7 of 2026-10-21
  9. VLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action Reuse 64 upvotes, #9 of 2026-10-21
  10. Learning the 4D World through Point Trajectory Forecasting 55 upvotes, #10 of 2026-10-21
  11. SolarRenderer: Few-Step Generative Reference Rendering 52 upvotes, #11 of 2026-10-21
  12. Correction Training for Diffusion Models 50 upvotes, #12 of 2026-10-21
  13. PhysSkin++: Real-Time and Generalizable Physics-Based Animation via Self-Supervised Neural Skinning 37 upvotes, #13 of 2026-10-21
  14. SWIFT: Self-guided Weak-to-strong dIFfusion models via Target extrapolation 35 upvotes, #14 of 2026-10-21
  15. MMEnterprise-Bench: Evaluating Enterprise Agents over Evolving Multimodal Records 33 upvotes, #15 of 2026-10-21
  16. SpecKCast: Spectral Kalman Forecasting for Diffusion Acceleration 29 upvotes, #16 of 2026-10-21
  17. EvaOmni: An Evidence-Grounded Benchmark for Audio-Visual Reasoning 28 upvotes, #17 of 2026-10-21
  18. Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative Offloading 27 upvotes, #18 of 2026-10-21
  19. Joints, Camera, Action! Co-Generating 3D Motion and Video for Arbitrary Rigs 27 upvotes, #18 of 2026-10-21
  20. EAU: Environment-Agent-User Co-evolution for Training User-Centric Agents 26 upvotes, #20 of 2026-10-21
  21. LLM2R: From Large Language Models to 3D Reassemblers 25 upvotes, #21 of 2026-10-21
  22. Epi: Towards a Persistent, Self-Improving Agent for Autonomous Research 24 upvotes, #22 of 2026-10-21
  23. Ask to Distill: Learning to Synthesize Data for On-Policy Distillation 24 upvotes, #22 of 2026-10-21
  24. FlowState: Execution State as Memory for Long-Horizon LLM Agents 23 upvotes, #24 of 2026-10-21
  25. Breaking Diffusion: Parallel Reasoning with Draft Refinement 23 upvotes, #24 of 2026-10-21
  26. Large Flow Language Models 23 upvotes, #24 of 2026-10-21
  27. Evolving Agent Interfaces 23 upvotes, #24 of 2026-10-21
  28. Painting Change Map as guidance for generating painting processes 20 upvotes, #28 of 2026-10-21
  29. Training the Whole Agent, Not Just the Model 20 upvotes, #28 of 2026-10-21
  30. Muse Glimmer 30B: An On-Device Agentic Model 19 upvotes, #30 of 2026-10-21
  31. Event-STVG: Spatio-Temporal Video Grounding for Long-Form Event Reasoning 18 upvotes, #31 of 2026-10-21
  32. From Short Clips to Real-Time Streams: Identity-Aware Supervision and Context-Efficient Generation for Video Face Swapping 18 upvotes, #31 of 2026-10-21
  33. EdgeKernelBench: Unified LLM-Driven Kernel Generation for Edge Backends 17 upvotes, #33 of 2026-10-21
  34. More Memory, Worse Agents: Error Reproduction and Anti-Persistence in LLM Agents 17 upvotes, #33 of 2026-10-21
  35. Good Judgment, Bad Advice: Decision-Model Controllers Raise Agent Cost Without Improving Success 16 upvotes, #35 of 2026-10-21
  36. HarnessRL: Co-Evolving Policies with a Just-in-Time Training Harness for Open-Ended RL 15 upvotes, #36 of 2026-10-21
  37. Harness Self-Evolution for Long-Horizon Terminal Agents 14 upvotes, #37 of 2026-10-21
  38. HybridResearcher: Bridging the Structured-Unstructured Boundary for Deep Research Agents 13 upvotes, #38 of 2026-10-21
  39. Unsupervised Tool Evolution for Coding Agents in Data-Lake Analytics 13 upvotes, #38 of 2026-10-21
  40. On the Limits of Scale: Inverse Scaling in Vision-Language Models 13 upvotes, #38 of 2026-10-21
  41. Self-Credit: Credit Assignment for Long-Horizon Agentic RL via Intra-group Rescaling 13 upvotes, #38 of 2026-10-21
  42. Evaluating Agentic Capabilities Before Post-Training: An End-to-End Approach 11 upvotes, #42 of 2026-10-21
  43. On the Fragility of Tasks under Composition: Implicit Budgeting in Language Models 11 upvotes, #42 of 2026-10-21
  44. PhysLDM: Latent Diffusion for High-Fidelity Deformable Simulation 10 upvotes, #44 of 2026-10-21
  45. PVBench: Quantifying Physical Plausibility in Video Generation with Object-Level Physical Measurements 10 upvotes, #44 of 2026-10-21
  46. Beyond Compilation: Evaluating Faithful Natural-Language-to-Lean Statement Formalization 9 upvotes, #46 of 2026-10-21
  47. λ_0: Scaling Open-World Egocentric Human Data for Humanoid Loco-Manipulation 8 upvotes, #47 of 2026-10-21
  48. WinARDet: Compressed Window Auto-Regression for Long-Horizon Vision-Language Grounding 8 upvotes, #47 of 2026-10-21
  49. On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs 8 upvotes, #47 of 2026-10-21
  50. Multimodal Reranking Is a System One Decision 8 upvotes, #47 of 2026-10-21
  51. EMBER: Predicting TCR--pMHC Binding Without a Docked Complex 7 upvotes, #51 of 2026-10-21
  52. OverSeer: Multi-Agent World Models Beyond What Agents See 6 upvotes, #52 of 2026-10-21
  53. OmniCam: Omni-Camera Trajectory Generation via Geometry-Grounded Pose Token Learning 5 upvotes, #53 of 2026-10-21
  54. ATEST: Adaptive Resource-Aware Test-Time Scaling for LLM Inference 5 upvotes, #53 of 2026-10-21
  55. Slot-JEPA: Object-Structured Prediction Interfaces for Latent World Models 5 upvotes, #53 of 2026-10-21
  56. HRoPE: Homography Relative Positional Encoding 5 upvotes, #53 of 2026-10-21
  57. HumanSkill-4D: Training-Free Language Querying of Animatable Gaussian Avatars in Space and Time 4 upvotes, #57 of 2026-10-21
  58. TRACE: Tracking Multimodal Clinical Trajectories for Inpatient Risk Prediction 4 upvotes, #57 of 2026-10-21
  59. Axiomatic World Modeling for Physics Reasoning 3 upvotes, #59 of 2026-10-21
  60. VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Token 3 upvotes, #59 of 2026-10-21
  61. What Makes a Good Symbol for Hybrid Discrete-Continuous Planning? 2 upvotes, #61 of 2026-10-21
  62. World Models with Configurable Perception 2 upvotes, #61 of 2026-10-21
  63. Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models 1 upvotes, #63 of 2026-10-21
  64. BiDex-100: A Comprehensive Benchmark for Bimanual Dexterous Manipulation Across 100+ Simulation Tasks 1 upvotes, #63 of 2026-10-21

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.