Daily Papers of 2025-08-25

  1. AgentFly: Fine-tuning LLM Agents without Fine-tuning LLMs 118 upvotes, #1 of 2025-08-25
  2. Beyond Pass@1: Self-Play with Variational Problem Synthesis Sustains RLVR 116 upvotes, #2 of 2025-08-25
  3. ODYSSEY: Open-World Quadrupeds Exploration and Manipulation for Long-Horizon Tasks 43 upvotes, #3 of 2025-08-25
  4. EgoTwin: Dreaming Body and View in First Person 18 upvotes, #4 of 2025-08-25
  5. AgentScope 1.0: A Developer-Centric Framework for Building Agentic Applications 15 upvotes, #5 of 2025-08-25
  6. CRISP: Persistent Concept Unlearning via Sparse Autoencoders 14 upvotes, #6 of 2025-08-25
  7. AetherCode: Evaluating LLMs' Ability to Win In Premier Programming Competitions 14 upvotes, #6 of 2025-08-25
  8. Selective Contrastive Learning for Weakly Supervised Affordance Grounding 12 upvotes, #8 of 2025-08-25
  9. End-to-End Agentic RAG System Training for Traceable Diagnostic Reasoning 9 upvotes, #9 of 2025-08-25
  10. Do What? Teaching Vision-Language-Action Models to Reject the Impossible 8 upvotes, #10 of 2025-08-25
  11. Distilled-3DGS:Distilled 3D Gaussian Splatting 7 upvotes, #11 of 2025-08-25
  12. TPLA: Tensor Parallel Latent Attention for Efficient Disaggregated Prefill \& Decode Inference 7 upvotes, #11 of 2025-08-25
  13. Sketch3DVE: Sketch-based 3D-Aware Scene Video Editing 6 upvotes, #13 of 2025-08-25
  14. Learnable SMPLify: A Neural Solution for Optimization-Free Human Pose Inverse Kinematics 2 upvotes, #14 of 2025-08-25
  15. CARFT: Boosting LLM Reasoning via Contrastive Learning with Annotated Chain-of-Thought-based Reinforced Fine-Tuning 2 upvotes, #14 of 2025-08-25
  16. InMind: Evaluating LLMs in Capturing and Applying Individual Human Reasoning Styles 2 upvotes, #14 of 2025-08-25
  17. Jailbreaking Commercial Black-Box LLMs with Explicitly Harmful Prompts 1 upvotes, #17 of 2025-08-25
  18. RotaTouille: Rotation Equivariant Deep Learning for Contours 1 upvotes, #17 of 2025-08-25

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.