Daily Papers of 2025-10-22

  1. Efficient Long-context Language Model Training by Core Attention Disaggregation 114 upvotes, #1 of 2025-10-22
  2. LightMem: Lightweight and Efficient Memory-Augmented Generation 105 upvotes, #2 of 2025-10-22
  3. World-in-World: World Models in a Closed-Loop World 83 upvotes, #3 of 2025-10-22
  4. UniGenBench++: A Unified Semantic Evaluation Benchmark for Text-to-Image Generation 66 upvotes, #4 of 2025-10-22
  5. DeepSeek-OCR: Contexts Optical Compression 63 upvotes, #5 of 2025-10-22
  6. Every Step Evolves: Scaling Reinforcement Learning for Trillion-Scale Thinking Model 61 upvotes, #6 of 2025-10-22
  7. Chem-R: Learning to Reason as a Chemist 51 upvotes, #7 of 2025-10-22
  8. MoGA: Mixture-of-Groups Attention for End-to-End Long Video Generation 38 upvotes, #8 of 2025-10-22
  9. Grasp Any Region: Towards Precise, Contextual Pixel Understanding for Multimodal LLMs 35 upvotes, #9 of 2025-10-22
  10. IF-VidCap: Can Video Caption Models Follow Instructions? 24 upvotes, #10 of 2025-10-22
  11. GAS: Improving Discretization of Diffusion ODEs via Generalized Adversarial Solver 20 upvotes, #11 of 2025-10-22
  12. Think with 3D: Geometric Imagination Grounded Spatial Reasoning from Limited Views 20 upvotes, #11 of 2025-10-22
  13. Towards Faithful and Controllable Personalization via Critique-Post-Edit Reinforcement Learning 19 upvotes, #13 of 2025-10-22
  14. MT-Video-Bench: A Holistic Video Understanding Benchmark for Evaluating Multimodal LLMs in Multi-Turn Dialogues 17 upvotes, #14 of 2025-10-22
  15. Is Multilingual LLM Watermarking Truly Multilingual? A Simple Back-Translation Solution 16 upvotes, #15 of 2025-10-22
  16. UltraGen: High-Resolution Video Generation with Hierarchical Attention 16 upvotes, #15 of 2025-10-22
  17. ssToken: Self-modulated and Semantic-aware Token Selection for LLM Fine-tuning 12 upvotes, #17 of 2025-10-22
  18. MUG-V 10B: High-efficiency Training Pipeline for Large Video Generation Models 9 upvotes, #18 of 2025-10-22
  19. ProCLIP: Progressive Vision-Language Alignment via LLM-based Embedder 9 upvotes, #18 of 2025-10-22
  20. AlphaQuanter: An End-to-End Tool-Orchestrated Agentic Reinforcement Learning Framework for Stock Trading 8 upvotes, #20 of 2025-10-22
  21. Video Reasoning without Training 7 upvotes, #21 of 2025-10-22
  22. DSI-Bench: A Benchmark for Dynamic Spatial Intelligence 7 upvotes, #21 of 2025-10-22
  23. Expanding the Action Space of LLMs to Reason Beyond Language 6 upvotes, #23 of 2025-10-22
  24. Extracting alignment data in open models 6 upvotes, #23 of 2025-10-22
  25. Unimedvl: Unifying Medical Multimodal Understanding And Generation Through Observation-Knowledge-Analysis 5 upvotes, #25 of 2025-10-22
  26. Planned Diffusion 5 upvotes, #25 of 2025-10-22
  27. PokeeResearch: Effective Deep Research via Reinforcement Learning from AI Feedback and Robust Reasoning Scaffold 4 upvotes, #27 of 2025-10-22
  28. When "Correct" Is Not Safe: Can We Trust Functionally Correct Patches Generated by Code Agents? 4 upvotes, #27 of 2025-10-22
  29. Mono4DGS-HDR: High Dynamic Range 4D Gaussian Splatting from Alternating-exposure Monocular Videos 4 upvotes, #27 of 2025-10-22
  30. Unleashing Scientific Reasoning for Bio-experimental Protocol Generation via Structured Component-based Reward Mechanism 3 upvotes, #30 of 2025-10-22
  31. PRISMM-Bench: A Benchmark of Peer-Review Grounded Multimodal Inconsistencies 3 upvotes, #30 of 2025-10-22
  32. Pruning Overparameterized Multi-Task Networks for Degraded Web Image Restoration 2 upvotes, #32 of 2025-10-22
  33. The Atomic Instruction Gap: Instruction-Tuned LLMs Struggle with Simple, Self-Contained Directives 2 upvotes, #32 of 2025-10-22
  34. Any-Depth Alignment: Unlocking Innate Safety Alignment of LLMs to Any-Depth 2 upvotes, #32 of 2025-10-22
  35. Static Sandboxes Are Inadequate: Modeling Societal Complexity Requires Open-Ended Co-Evolution in LLM-Based Multi-Agent Simulations 1 upvotes, #35 of 2025-10-22
  36. Predicting the Unpredictable: Reproducible BiLSTM Forecasting of Incident Counts in the Global Terrorism Database (GTD) 1 upvotes, #35 of 2025-10-22
  37. EvoSyn: Generalizable Evolutionary Data Synthesis for Verifiable Learning 1 upvotes, #35 of 2025-10-22

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.