Daily Papers of 2026-09-22

  1. OmniEdu: Open Foundation Models for Learning and Teaching 233 upvotes, #1 of 2026-09-22
  2. RRSI: Regularized Recursive Self-Improvement of Agent Harnesses 216 upvotes, #2 of 2026-09-22
  3. WorldCrafter: Consistent Video World Model with Implicit 3D-aware Memory 156 upvotes, #3 of 2026-09-22
  4. GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay 130 upvotes, #4 of 2026-09-22
  5. Transferring the Intelligence of VLMs to Robotic Control 118 upvotes, #5 of 2026-09-22
  6. Grounded Action Model: 3D Grounding as a Foundation for Robotics 89 upvotes, #6 of 2026-09-22
  7. VideoGen-Agent: Reinforcing Video Generation Agents 71 upvotes, #7 of 2026-09-22
  8. Document Retrieval-Aware Chunking (D-RAC): Universal Retrieval-Aware Ingestion of Enterprise Documents via PDF Normalization and Multimodal Markdown Conversion 68 upvotes, #8 of 2026-09-22
  9. onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction 55 upvotes, #9 of 2026-09-22
  10. One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering Agents 50 upvotes, #10 of 2026-09-22
  11. Harness-Zero: Harness Distillation via Agent-as-Harness 37 upvotes, #11 of 2026-09-22
  12. Why Do Video Diffusion Models Violate Physics? Unveiling the Flaws in Attention Mechanisms 30 upvotes, #12 of 2026-09-22
  13. Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents 28 upvotes, #13 of 2026-09-22
  14. CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies 28 upvotes, #13 of 2026-09-22
  15. Deep Persona: A Psychologically Grounded Architecture and Evaluation Framework for Role-Playing Agents and Simulations 22 upvotes, #15 of 2026-09-22
  16. HuRo: Robotizing Human Videos for Scalable VLA Pretraining 18 upvotes, #16 of 2026-09-22
  17. ShieldVLA: Feasibility-Aware Safety Alignment for Vision-Language-Action Models 18 upvotes, #16 of 2026-09-22
  18. Towards Full Pipeline FP8 Reinforcement Learning for LLMs 17 upvotes, #18 of 2026-09-22
  19. 1% of Tokens Can Be Enough: On Gradient Estimation in On-Policy Distillation 16 upvotes, #19 of 2026-09-22
  20. The Functionalizer: Lossless Functional Decomposition for Subword Tokenization 14 upvotes, #20 of 2026-09-22
  21. ACLArena: Agent Continue Learning in Multi-stage Post-training 12 upvotes, #21 of 2026-09-22
  22. Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies 12 upvotes, #21 of 2026-09-22
  23. Streaming Video Editing with Easy Adaptation 11 upvotes, #23 of 2026-09-22
  24. Complex KDA: Understanding and Enhancing the Expressivity of Kimi Delta Attention 11 upvotes, #23 of 2026-09-22
  25. SkillSpec: Intent-Masked Specification Reasoning for Agent Skill Correctness 9 upvotes, #25 of 2026-09-22
  26. A Lie Detector Test for Language Models: Reading Knowledge a Model Won't Reveal 9 upvotes, #25 of 2026-09-22
  27. Measuring the Checker: Mutation Analysis for GPU-Kernel Benchmark Oracles 7 upvotes, #27 of 2026-09-22
  28. The information geometry of large language models is shared, learned, and controllable 6 upvotes, #28 of 2026-09-22
  29. Mira-Scene: Pixel-Aligned Layouts for Generative 3D Scene 5 upvotes, #29 of 2026-09-22
  30. EDGEGEN: Improving Tool-Calling Agents Beyond Happy Paths with Synthetic Edge Case Generation 5 upvotes, #29 of 2026-09-22
  31. Prediction-Powered Smoothing and Validation for Disaggregated AI Evaluation 4 upvotes, #31 of 2026-09-22
  32. UltraTex: Unleashing 2K Multi-View Diffusion for 3D Texturing 3 upvotes, #32 of 2026-09-22
  33. TAPe+ML: A Compact Structured Representation for Multi-Task Computer Vision 2 upvotes, #33 of 2026-09-22

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.