Daily Papers of 2025-07-23

  1. Beyond Context Limits: Subconscious Threads for Long-Horizon Reasoning 112 upvotes, #1 of 2025-07-23
  2. Step-Audio 2 Technical Report 58 upvotes, #2 of 2025-07-23
  3. MegaScience: Pushing the Frontiers of Post-Training Datasets for Science Reasoning 48 upvotes, #3 of 2025-07-23
  4. Upsample What Matters: Region-Adaptive Latent Sampling for Accelerated Diffusion Transformers 35 upvotes, #4 of 2025-07-23
  5. ThinkAct: Vision-Language-Action Reasoning via Reinforced Visual Latent Planning 35 upvotes, #4 of 2025-07-23
  6. Zebra-CoT: A Dataset for Interleaved Vision Language Reasoning 32 upvotes, #6 of 2025-07-23
  7. Semi-off-Policy Reinforcement Learning for Vision-Language Slow-thinking Reasoning 22 upvotes, #7 of 2025-07-23
  8. Experience is the Best Teacher: Grounding VLMs for Robotics through Self-Generated Memory 21 upvotes, #8 of 2025-07-23
  9. RefCritic: Training Long Chain-of-Thought Critic Models with Refinement Feedback 13 upvotes, #9 of 2025-07-23
  10. HOComp: Interaction-Aware Human-Object Composition 12 upvotes, #10 of 2025-07-23
  11. SPAR: Scholar Paper Retrieval with LLM-based Agents for Enhanced Academic Search 11 upvotes, #11 of 2025-07-23
  12. Task-Specific Zero-shot Quantization-Aware Training for Object Detection 9 upvotes, #12 of 2025-07-23
  13. PrefPalette: Personalized Preference Modeling with Latent Attributes 8 upvotes, #13 of 2025-07-23
  14. ObjectGS: Object-aware Scene Reconstruction and Scene Understanding via Gaussian Splatting 7 upvotes, #14 of 2025-07-23
  15. Does More Inference-Time Compute Really Help Robustness? 6 upvotes, #15 of 2025-07-23
  16. Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuning 2 upvotes, #16 of 2025-07-23

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.