Daily Papers of 2025-05-09

  1. Perception, Reason, Think, and Plan: A Survey on Large Multimodal Reasoning Models 134 upvotes, #1 of 2025-05-09
  2. Flow-GRPO: Training Flow Matching Models via Online RL 68 upvotes, #2 of 2025-05-09
  3. Sentient Agent as a Judge: Evaluating Higher-Order Social Cognition in Large Language Models 25 upvotes, #3 of 2025-05-09
  4. Generating Physically Stable and Buildable LEGO Designs from Text 24 upvotes, #4 of 2025-05-09
  5. Scalable Chain of Thoughts via Elastic Reasoning 23 upvotes, #5 of 2025-05-09
  6. FG-CLIP: Fine-Grained Visual and Textual Alignment 16 upvotes, #6 of 2025-05-09
  7. 3D Scene Generation: A Survey 16 upvotes, #6 of 2025-05-09
  8. X-Reasoner: Towards Generalizable Reasoning Across Modalities and Domains 14 upvotes, #8 of 2025-05-09
  9. StreamBridge: Turning Your Offline Video Large Language Model into a Proactive Streaming Assistant 13 upvotes, #9 of 2025-05-09
  10. Putting the Value Back in RL: Better Test-Time Scaling by Unifying LLM Reasoners With Verifiers 12 upvotes, #10 of 2025-05-09
  11. PlaceIt3D: Language-Guided Object Placement in Real 3D Scenes 11 upvotes, #11 of 2025-05-09
  12. ICon: In-Context Contribution for Automatic Data Selection 11 upvotes, #11 of 2025-05-09
  13. Chain-of-Thought Tokens are Computer Program Variables 8 upvotes, #13 of 2025-05-09
  14. WaterDrum: Watermarking for Data-centric Unlearning Metric 8 upvotes, #13 of 2025-05-09
  15. Crosslingual Reasoning through Test-Time Scaling 8 upvotes, #13 of 2025-05-09
  16. LiftFeat: 3D Geometry-Aware Local Feature Matching 7 upvotes, #16 of 2025-05-09
  17. Vision-Language-Action Models: Concepts, Progress, Applications and Challenges 7 upvotes, #16 of 2025-05-09
  18. SIMPLEMIX: Frustratingly Simple Mixing of Off- and On-policy Data in Language Model Preference Learning 6 upvotes, #18 of 2025-05-09
  19. BrowseComp-ZH: Benchmarking Web Browsing Ability of Large Language Models in Chinese 4 upvotes, #19 of 2025-05-09

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.