Daily Papers of 2026-09-04

  1. Compile by Training: Turning Natural-Language Specifications into Local Neural Functions 380 upvotes, #1 of 2026-09-04
  2. Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments 291 upvotes, #2 of 2026-09-04
  3. LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes 233 upvotes, #3 of 2026-09-04
  4. Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning 177 upvotes, #4 of 2026-09-04
  5. Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training 153 upvotes, #5 of 2026-09-04
  6. LatentPress: Context Compression Beyond Text and Vision 118 upvotes, #6 of 2026-09-04
  7. RoboTok: An Internet-Scale Data Engine for Human Demonstration Retrieval and Dexterous Manipulation Learning 116 upvotes, #7 of 2026-09-04
  8. Rethinking On-Policy Distillation of Large Language Models II: One Training Example 91 upvotes, #8 of 2026-09-04
  9. Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM 80 upvotes, #9 of 2026-09-04
  10. Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States 66 upvotes, #10 of 2026-09-04
  11. Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction 46 upvotes, #11 of 2026-09-04
  12. Editable Visual Design 42 upvotes, #12 of 2026-09-04
  13. Last Translation Benchmark 37 upvotes, #13 of 2026-09-04
  14. The Missing Temporal Link: Temporal Context Routing for Script-Driven Audio-Video Generation 35 upvotes, #14 of 2026-09-04
  15. RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests 32 upvotes, #15 of 2026-09-04
  16. Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding 31 upvotes, #16 of 2026-09-04
  17. WorldReward: Reward Modeling for Camera-Conditioned World Models 26 upvotes, #17 of 2026-09-04
  18. CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation 26 upvotes, #17 of 2026-09-04
  19. DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training 26 upvotes, #17 of 2026-09-04
  20. PACE: Towards Surfacing Hidden Conflicts in User Requests 24 upvotes, #20 of 2026-09-04
  21. Environment Evolution for Terminal Agents 22 upvotes, #21 of 2026-09-04
  22. Using Grounded Theory for Agent Behavior Analysis at Scale 19 upvotes, #22 of 2026-09-04
  23. FlashRender: Few-Step Generative Rendering via Camera-Controlled Video MeanFlow 19 upvotes, #22 of 2026-09-04
  24. Select, Compress, Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLMs 16 upvotes, #24 of 2026-09-04
  25. Principia: Relational Physics Tests for Video Models 16 upvotes, #24 of 2026-09-04
  26. QCell: Recombining and Aligning Cell Queries for Overlapping Instance Segmentation 14 upvotes, #26 of 2026-09-04
  27. Percolation Dynamics in Optimization : Variance Cascades and Discrete Scale Invariance 14 upvotes, #26 of 2026-09-04
  28. Let Confidence Change, Not the Prediction: Prediction-Preserving Repair for Post-hoc Calibration 13 upvotes, #28 of 2026-09-04
  29. VeriPhy: Agentic Physical Reasoning for World Model Evaluation and Refinement 13 upvotes, #28 of 2026-09-04
  30. Locked at the Entrance, Open Inside: Where RLVR Narrows the Solution Space 10 upvotes, #30 of 2026-09-04
  31. A Common Measure of Communication for Speech Brain-Computer Interfaces 10 upvotes, #30 of 2026-09-04

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.