Daily Papers of 2026-10-28

  1. The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents 539 upvotes, #1 of 2026-10-28
  2. Multiscale Policy Optimization for Long-Horizon LLM Agents 376 upvotes, #2 of 2026-10-28
  3. ChronoReplay: One Answer Interface Is Not Enough for Streaming Spatial Intelligence 366 upvotes, #3 of 2026-10-28
  4. VibeVideoBench: Evaluating Multimodal Agents on Code-Driven Video Generation 173 upvotes, #4 of 2026-10-28
  5. Memory Tax: Dissecting the Net Value of Agent Memory 147 upvotes, #5 of 2026-10-28
  6. PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations 141 upvotes, #6 of 2026-10-28
  7. Hy-Nomas: Towards Free-form Interactive World Generation 108 upvotes, #7 of 2026-10-28
  8. Can Discrete Tokens Serve as Semantically Lossless Representations of Visual Signals? 83 upvotes, #8 of 2026-10-28
  9. HABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent Memory 80 upvotes, #9 of 2026-10-28
  10. Rollback: Experience Backtracking in World Models for Policy Improvement 66 upvotes, #10 of 2026-10-28
  11. ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning 51 upvotes, #11 of 2026-10-28
  12. CodeCap-Bench: Can Code Serve as a Good Video Caption? 48 upvotes, #12 of 2026-10-28
  13. LociArt: A Unified Paradigm for Free-Style Local Video Stylization 38 upvotes, #13 of 2026-10-28
  14. Zooming without Zooming v2: From Seeing Details to Connecting Evidence for Generalizable Multimodal Perception 24 upvotes, #14 of 2026-10-28
  15. RSI Arena: Towards Recursive Self-Improvement with Human Feedback 22 upvotes, #15 of 2026-10-28
  16. EmbeddingGemma 2: Unified and Efficient Multimodal Representations 20 upvotes, #16 of 2026-10-28
  17. MatAnyone-E: Sparse Memory Propagation for Efficient High-Resolution Video Matting 18 upvotes, #17 of 2026-10-28
  18. h1: A Taste of the Bitter Lesson in Automated Theorem Proving 16 upvotes, #18 of 2026-10-28
  19. Omni-InteractScope: Benchmarking Proactive and Agentic Interaction in Continuous Omnimodal Streams 14 upvotes, #19 of 2026-10-28
  20. On the SPOT: Enhancing Speculative Decoding for Diffusion Language Models via Commit-Order Prediction 12 upvotes, #20 of 2026-10-28
  21. Channel Selection for Mixture-of-Experts Decoding Needs No Calibration Data 12 upvotes, #20 of 2026-10-28
  22. What Effort Level Should I Choose? On the Economics of Reasoning Effort in LLM Agents 11 upvotes, #22 of 2026-10-28
  23. Nemotron-Labs-Diffusion-Uni: Elastic Self-Refining Unified Diffusion LLM with Cross-Paradigm On-Policy Distillation 11 upvotes, #22 of 2026-10-28
  24. Cross-modal Steering Vectors Exist Across Both Image and Text Generation 10 upvotes, #24 of 2026-10-28
  25. The Lexical Imprint of Pre-training Data 9 upvotes, #25 of 2026-10-28
  26. RynnWorld-Latent: A Cross-Embodiment World Model Conditioned on Universal Latent Actions 9 upvotes, #25 of 2026-10-28
  27. Benchmarking Obstacle-Conditioned Safety in Household Robot Navigation 8 upvotes, #27 of 2026-10-28
  28. SearchHorizonBench: A Benchmark for Search Value in AI Agents 7 upvotes, #28 of 2026-10-28
  29. Your Policy Is Secretly a Harness: The Decision Geometry of Agent Interfaces 7 upvotes, #28 of 2026-10-28
  30. Can LLMs Uncover Hidden Physical Laws through Free Exploration? 7 upvotes, #28 of 2026-10-28
  31. OvisOCR3 Technical Report 6 upvotes, #31 of 2026-10-28
  32. Harness-Model Collaboration for Long-Horizon Continual Learning 6 upvotes, #31 of 2026-10-28
  33. Timestral 6 upvotes, #31 of 2026-10-28
  34. Learning and Reasoning with Geometry: How Generation Helps Spatial Understanding 5 upvotes, #34 of 2026-10-28
  35. Vision-Language-Action or World-Action? Just Action Models 5 upvotes, #34 of 2026-10-28
  36. Rare, Relevant, and Still Wrong: How Memory Fails LLM Group Decisions 5 upvotes, #34 of 2026-10-28
  37. What Outcome RL Buys: Finishing, Reliability, Rarely Reach 4 upvotes, #37 of 2026-10-28
  38. How Much Does the Hint Matter? Scope Conditions for Hindsight Self-Distillation in Financial Reasoning 4 upvotes, #37 of 2026-10-28
  39. MLE-World: Predicting Machine Learning Experiment Outcomes with a Learned World Model 4 upvotes, #37 of 2026-10-28
  40. Sub-monolayer tracking of molecular states at buried interfaces 3 upvotes, #40 of 2026-10-28
  41. Beyond the Estimate: Wiener-Informed Decoding for Easier Training of Pixel-Space Flow Models 3 upvotes, #40 of 2026-10-28
  42. Quaternion Approximate Networks for Enhanced Image Classification and Oriented Object Detection 2 upvotes, #42 of 2026-10-28
  43. AlphaBet: Reinforcement Learning for LLM Event Forecasting Agents 2 upvotes, #42 of 2026-10-28
  44. Truncate or Not? A Scaling Study of Looped Transformers with Truncated Backpropagation 2 upvotes, #42 of 2026-10-28
  45. MoveBench: A Rubric-Based Benchmark for Evaluating Large Language Models in Sports and Health Coaching 2 upvotes, #42 of 2026-10-28
  46. T-DIG: Text-Conditioned Diversity for Image Generation 2 upvotes, #42 of 2026-10-28
  47. Self-Assessment Reinforcement Learning for Search-State Awareness in Retrieval-Augmented Large Language Models 2 upvotes, #42 of 2026-10-28
  48. PhysEvolve: Verifier-grounded Self-evolution for Physical Reasoning with Executable Worlds 2 upvotes, #42 of 2026-10-28
  49. AdaPRA: Process Reward Agents with Adaptive Rubrics for Data Science 1 upvotes, #49 of 2026-10-28
  50. Lipi-Ghor: A Large-Scale, Diarized, Multi-Domain Bengali Speech Corpus and a Scalable Pipeline for Building It 1 upvotes, #49 of 2026-10-28

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.