Daily Papers of 2026-09-25

  1. Training Object Permanence in World Models 238 upvotes, #1 of 2026-09-25
  2. Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs 100 upvotes, #2 of 2026-09-25
  3. WanPE: Towards Cinematic Prompt Enhancement for Modern Text-to-Video Generation 46 upvotes, #3 of 2026-09-25
  4. Agent-Editing World Model: Rethinking World Modeling for LLM Agents 42 upvotes, #4 of 2026-09-25
  5. Rufus-Air: An Open LLM Post-Training Recipe 37 upvotes, #5 of 2026-09-25
  6. OmniEcho: Spatial Audio Understanding for Embodied Agents 35 upvotes, #6 of 2026-09-25
  7. Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents 32 upvotes, #7 of 2026-09-25
  8. ExplorationBench: Measuring AI Systems' Exploration in Verifiable Alien Worlds 28 upvotes, #8 of 2026-09-25
  9. PUBG Ally: A Conversational Embodied Agent as an AI Teammate 25 upvotes, #9 of 2026-09-25
  10. Coding Agents for Generalized Task and Motion Planning Problems 25 upvotes, #9 of 2026-09-25
  11. Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures 24 upvotes, #11 of 2026-09-25
  12. IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis 20 upvotes, #12 of 2026-09-25
  13. DeltaWAM: Delta World Action Models for Bimanual Manipulation 19 upvotes, #13 of 2026-09-25
  14. Parts-of-Speech as Emergent Categories in SAE Latent Space 19 upvotes, #13 of 2026-09-25
  15. Learning to Discover Interesting Mathematics 15 upvotes, #15 of 2026-09-25
  16. RGBD20K: A Large-Scale Benchmark for RGB-D Semantic Segmentation 15 upvotes, #15 of 2026-09-25
  17. AgentKernel: The Trust-Native Agentic Operating System 15 upvotes, #15 of 2026-09-25
  18. World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal 15 upvotes, #15 of 2026-09-25
  19. AV-GRPO: Modality-Anchored Decoupling Diffusion Reinforcement Learning for Joint Audio-Video Generation 12 upvotes, #19 of 2026-09-25
  20. Neural Spectral Capacity: Measuring and Designing Architectures from Network Specification Alone 11 upvotes, #20 of 2026-09-25
  21. ViRDM: Taming Representation Distribution Matching for Few-Step Causal Video Generation 10 upvotes, #21 of 2026-09-25
  22. Rate-distortion optimization for full-reference image quality metrics via stochastic Hessian estimates 6 upvotes, #22 of 2026-09-25

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.