Daily Papers of 2025-10-23

  1. Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning 100 upvotes, #1 of 2025-10-23
  2. BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clipping 80 upvotes, #2 of 2025-10-23
  3. LoongRL:Reinforcement Learning for Advanced Reasoning over Long Contexts 58 upvotes, #3 of 2025-10-23
  4. Language Models are Injective and Hence Invertible 57 upvotes, #4 of 2025-10-23
  5. Attention Sinks in Diffusion Language Models 47 upvotes, #5 of 2025-10-23
  6. GigaBrain-0: A World Model-Powered Vision-Language-Action Model 42 upvotes, #6 of 2025-10-23
  7. Pico-Banana-400K: A Large-Scale Dataset for Text-Guided Image Editing 27 upvotes, #7 of 2025-10-23
  8. Unified Reinforcement and Imitation Learning for Vision-Language Models 26 upvotes, #8 of 2025-10-23
  9. HSCodeComp: A Realistic and Expert-level Benchmark for Deep Search Agents in Hierarchical Rule Application 26 upvotes, #8 of 2025-10-23
  10. DeepWideSearch: Benchmarking Depth and Width in Agentic Information Seeking 26 upvotes, #8 of 2025-10-23
  11. VideoAgentTrek: Computer Use Pretraining from Unlabeled Videos 19 upvotes, #11 of 2025-10-23
  12. DaMo: Data Mixing Optimizer in Fine-tuning Multimodal LLMs for Mobile Phone Agents 16 upvotes, #12 of 2025-10-23
  13. DeLeaker: Dynamic Inference-Time Reweighting For Semantic Leakage Mitigation in Text-to-Image Models 10 upvotes, #13 of 2025-10-23
  14. Directional Reasoning Injection for Fine-Tuning MLLMs 10 upvotes, #13 of 2025-10-23
  15. Decomposed Attention Fusion in MLLMs for Training-Free Video Reasoning Segmentation 10 upvotes, #13 of 2025-10-23
  16. olmOCR 2: Unit Test Rewards for Document OCR 10 upvotes, #13 of 2025-10-23
  17. KORE: Enhancing Knowledge Injection for Large Multimodal Models via Knowledge-Oriented Augmentations and Constraints 9 upvotes, #17 of 2025-10-23
  18. OmniNWM: Omniscient Driving Navigation World Models 8 upvotes, #18 of 2025-10-23
  19. FinSight: Towards Real-World Financial Deep Research 7 upvotes, #19 of 2025-10-23
  20. ProfBench: Multi-Domain Rubrics requiring Professional Knowledge to Answer and Judge 7 upvotes, #19 of 2025-10-23
  21. Steering Autoregressive Music Generation with Recursive Feature Machines 7 upvotes, #19 of 2025-10-23
  22. ColorAgent: Building A Robust, Personalized, and Interactive OS Agent 7 upvotes, #19 of 2025-10-23
  23. From Charts to Code: A Hierarchical Benchmark for Multimodal Models 6 upvotes, #23 of 2025-10-23
  24. Are they lovers or friends? Evaluating LLMs' Social Reasoning in English and Korean Dialogues 6 upvotes, #23 of 2025-10-23
  25. MINED: Probing and Updating with Multimodal Time-Sensitive Knowledge for Large Multimodal Models 6 upvotes, #23 of 2025-10-23
  26. NeuroAda: Activating Each Neuron's Potential for Parameter-Efficient Fine-Tuning 5 upvotes, #26 of 2025-10-23
  27. TheMCPCompany: Creating General-purpose Agents with Task-specific Tools 5 upvotes, #26 of 2025-10-23
  28. Accelerating Vision Transformers with Adaptive Patch Sizes 4 upvotes, #28 of 2025-10-23
  29. RIR-Mega: a large-scale simulated room impulse response dataset for machine learning and room acoustics modeling 4 upvotes, #28 of 2025-10-23
  30. Text or Pixels? It Takes Half: On the Token Efficiency of Visual Text Inputs in Multimodal LLMs 3 upvotes, #30 of 2025-10-23
  31. AlphaOPT: Formulating Optimization Programs with Self-Improving LLM Experience Library 3 upvotes, #30 of 2025-10-23
  32. See the Text: From Tokenization to Visual Reading 3 upvotes, #30 of 2025-10-23
  33. Learning from the Best, Differently: A Diversity-Driven Rethinking on Data Selection 3 upvotes, #30 of 2025-10-23
  34. What Questions Should Robots Be Able to Answer? A Dataset of User Questions for Explainable Robotics 2 upvotes, #34 of 2025-10-23
  35. Machine Text Detectors are Membership Inference Attacks 2 upvotes, #34 of 2025-10-23
  36. When Do Transformers Learn Heuristics for Graph Connectivity? 2 upvotes, #34 of 2025-10-23
  37. SAVANT: Semantic Analysis with Vision-Augmented Anomaly deTection 4 upvotes, #37 of 2025-10-23

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.