Daily Papers of 2025-08-28

  1. Beyond Transcription: Mechanistic Interpretability in ASR 83 upvotes, #1 of 2025-08-28
  2. Self-Rewarding Vision-Language Model via Reasoning Decomposition 77 upvotes, #2 of 2025-08-28
  3. CODA: Coordinating the Cerebrum and Cerebellum for a Dual-Brain Computer Use Agent with Decoupled Reinforcement Learning 35 upvotes, #3 of 2025-08-28
  4. Analysing Chain of Thought Dynamics: Active Guidance or Unfaithful Post-hoc Rationalisation? 29 upvotes, #4 of 2025-08-28
  5. Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies 28 upvotes, #5 of 2025-08-28
  6. MIDAS: Multimodal Interactive Digital-human Synthesis via Real-time Autoregressive Video Generation 27 upvotes, #6 of 2025-08-28
  7. Diffusion Language Models Know the Answer Before Decoding 22 upvotes, #7 of 2025-08-28
  8. Predicting the Order of Upcoming Tokens Improves Language Modeling 20 upvotes, #8 of 2025-08-28
  9. AudioStory: Generating Long-Form Narrative Audio with Large Language Models 20 upvotes, #8 of 2025-08-28
  10. StepWiser: Stepwise Generative Judges for Wiser Reasoning 19 upvotes, #10 of 2025-08-28
  11. Gaze into the Heart: A Multi-View Video Dataset for rPPG and Health Biomarkers Estimation 14 upvotes, #11 of 2025-08-28
  12. Mind the Third Eye! Benchmarking Privacy Awareness in MLLM-powered Smartphone Agents 11 upvotes, #12 of 2025-08-28
  13. SEAM: Semantically Equivalent Across Modalities Benchmark for Vision-Language Models 9 upvotes, #13 of 2025-08-28
  14. MotionFlux: Efficient Text-Guided Motion Generation through Rectified Flow Matching and Preference Alignment 9 upvotes, #13 of 2025-08-28
  15. DeepScholar-Bench: A Live Benchmark and Automated Evaluation for Generative Research Synthesis 7 upvotes, #15 of 2025-08-28
  16. Taming the Chaos: Coordinated Autoscaling for Heterogeneous and Disaggregated LLM Inference 4 upvotes, #16 of 2025-08-28
  17. Training a Foundation Model for Materials on a Budget 2 upvotes, #17 of 2025-08-28

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.