Daily Papers of 2026-01-29

  1. Advancing Open-source World Models 116 upvotes, #1 of 2026-01-29
  2. Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question Reformulation 116 upvotes, #1 of 2026-01-29
  3. Innovator-VL: A Multimodal Large Language Model for Scientific Discovery 76 upvotes, #3 of 2026-01-29
  4. DeepSeek-OCR 2: Visual Causal Flow 53 upvotes, #4 of 2026-01-29
  5. Reinforcement Learning via Self-Distillation 36 upvotes, #5 of 2026-01-29
  6. Spark: Strategic Policy-Aware Exploration via Dynamic Branching for Long-Horizon Agentic Learning 22 upvotes, #6 of 2026-01-29
  7. Linear representations in language models can change dramatically over a conversation 21 upvotes, #7 of 2026-01-29
  8. AACR-Bench: Evaluating Automatic Code Review with Holistic Repository-Level Context 15 upvotes, #8 of 2026-01-29
  9. SERA: Soft-Verified Efficient Repository Agents 11 upvotes, #9 of 2026-01-29
  10. Group Distributionally Robust Optimization-Driven Reinforcement Learning for LLM Reasoning 9 upvotes, #10 of 2026-01-29
  11. How AI Impacts Skill Formation 8 upvotes, #11 of 2026-01-29
  12. OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task Execution 8 upvotes, #11 of 2026-01-29
  13. FP8-RL: A Practical and Stable Low-Precision Stack for LLM Reinforcement Learning 6 upvotes, #13 of 2026-01-29
  14. VERGE: Formal Refinement and Guidance Engine for Verifiable LLM Reasoning 6 upvotes, #13 of 2026-01-29
  15. Training Reasoning Models on Saturated Problems via Failure-Prefix Conditioning 5 upvotes, #15 of 2026-01-29
  16. UPLiFT: Efficient Pixel-Dense Feature Upsampling with Local Attenders 4 upvotes, #16 of 2026-01-29
  17. SE-DiCoW: Self-Enrolled Diarization-Conditioned Whisper 3 upvotes, #17 of 2026-01-29
  18. RIR-Mega-Speech: A Reverberant Speech Corpus with Comprehensive Acoustic Metadata and Reproducible Evaluation 3 upvotes, #17 of 2026-01-29
  19. Persona Prompting as a Lens on LLM Social Reasoning 3 upvotes, #17 of 2026-01-29
  20. Shallow-π: Knowledge Distillation for Flow-based VLAs 2 upvotes, #20 of 2026-01-29
  21. GDCNet: Generative Discrepancy Comparison Network for Multimodal Sarcasm Detection 2 upvotes, #20 of 2026-01-29
  22. SketchDynamics: Exploring Free-Form Sketches for Dynamic Intent Expression in Animation Generation 1 upvotes, #22 of 2026-01-29

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.