Daily Papers of 2026-09-16

  1. LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence 660 upvotes, #1 of 2026-09-16
  2. Continual Learning Mechanisms Compose for Long-Horizon Memorization 372 upvotes, #2 of 2026-09-16
  3. ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement 210 upvotes, #3 of 2026-09-16
  4. AI for Games in the Foundation Model Era 145 upvotes, #4 of 2026-09-16
  5. StepAudio 3 Realtime Technical Report 125 upvotes, #5 of 2026-09-16
  6. StepAudio 3 Music Technical Report 107 upvotes, #6 of 2026-09-16
  7. Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement 82 upvotes, #7 of 2026-09-16
  8. ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agents 30 upvotes, #8 of 2026-09-16
  9. HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness 21 upvotes, #9 of 2026-09-16
  10. The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement 17 upvotes, #10 of 2026-09-16
  11. FLAT: Resampling Image and Text into 1D Flexible-Length Aligned Transmodal Tokens for Retrieval and Generation 16 upvotes, #11 of 2026-09-16
  12. Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models 13 upvotes, #12 of 2026-09-16
  13. Another Blueprint In The Wall: How to Ask Frontier AI Like a Kid? 12 upvotes, #13 of 2026-09-16
  14. ImpossibleRubrics: Stress-Testing Generated Rubrics as Reward Signals 11 upvotes, #14 of 2026-09-16
  15. Modality-Autoregressive World-Action Models 11 upvotes, #14 of 2026-09-16
  16. PhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion Control 10 upvotes, #16 of 2026-09-16
  17. Training Specialist Models without Reasoning Trajectories for Domain Expert Distillation 9 upvotes, #17 of 2026-09-16
  18. Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States 9 upvotes, #17 of 2026-09-16
  19. Disentangling Representation Evolution in Transformers through Directional Decomposition 9 upvotes, #17 of 2026-09-16
  20. The Router Within: Eliciting Native Skill Routing from a Frozen LLM 7 upvotes, #20 of 2026-09-16
  21. Register Tokens for Bounded-State Reasoning in Diffusion Language Models 7 upvotes, #20 of 2026-09-16
  22. RelateAnything: Real-Time Open-Vocabulary Relation Prediction From Any Inputs 4 upvotes, #22 of 2026-09-16
  23. Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration 4 upvotes, #22 of 2026-09-16
  24. Convergent Emergence of In-Context Learning Across Modalities 3 upvotes, #24 of 2026-09-16
  25. OmniHarness: Harnessing Generalizable Visual Generation via Symbolic Policy Learning 3 upvotes, #24 of 2026-09-16
  26. Emergence World: Adversarial Stress-Testing of Long-Horizon Multi-Agent Systems 3 upvotes, #24 of 2026-09-16

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.