Daily Papers of 2025-07-15

  1. Reasoning or Memorization? Unreliable Results of Reinforcement Learning Due to Data Contamination 77 upvotes, #1 of 2025-07-15
  2. Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation 57 upvotes, #2 of 2025-07-15
  3. SpeakerVid-5M: A Large-Scale High-Quality Dataset for Audio-Visual Dyadic Interactive Human Generation 48 upvotes, #3 of 2025-07-15
  4. EmbRACE-3K: Embodied Reasoning and Action in Complex Environments 33 upvotes, #4 of 2025-07-15
  5. REST: Stress Testing Large Reasoning Models by Asking Multiple Problems at Once 28 upvotes, #5 of 2025-07-15
  6. MoVieS: Motion-Aware 4D Dynamic View Synthesis in One Second 22 upvotes, #6 of 2025-07-15
  7. LayerCake: Token-Aware Contrastive Decoding within Large Language Model Layers 20 upvotes, #7 of 2025-07-15
  8. From KMMLU-Redux to KMMLU-Pro: A Professional Korean Benchmark Suite for LLM Evaluation 16 upvotes, #8 of 2025-07-15
  9. CompassJudger-2: Towards Generalist Judge Model via Verifiable Rewards 16 upvotes, #8 of 2025-07-15
  10. Subject-Consistent and Pose-Diverse Text-to-Image Generation 15 upvotes, #10 of 2025-07-15
  11. DreamPoster: A Unified Framework for Image-Conditioned Generative Poster Design 11 upvotes, #11 of 2025-07-15
  12. A Practical Two-Stage Recipe for Mathematical LLMs: Maximizing Accuracy with SFT and Efficiency with Reinforcement Learning 10 upvotes, #12 of 2025-07-15
  13. Favicon Trojans: Executable Steganography Via Ico Alpha Channel Exploitation 6 upvotes, #13 of 2025-07-15
  14. Sound and Complete Neuro-symbolic Reasoning with LLM-Grounded Interpretations 2 upvotes, #14 of 2025-07-15
  15. Hashed Watermark as a Filter: Defeating Forging and Overwriting Attacks in Weight-based Neural Network Watermarking 2 upvotes, #14 of 2025-07-15

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.