Daily Papers of 2026-03-09

  1. Penguin-VL: Exploring the Efficiency Limits of VLM with LLM-based Vision Encoders 105 upvotes, #1 of 2026-03-09
  2. BandPO: Bridging Trust Regions and Ratio Clipping via Probability-Aware Bounds for LLM Reinforcement Learning 54 upvotes, #2 of 2026-03-09
  3. Planning in 8 Tokens: A Compact Discrete Tokenizer for Latent World Model 36 upvotes, #3 of 2026-03-09
  4. WildActor: Unconstrained Identity-Preserving Video Generation 35 upvotes, #4 of 2026-03-09
  5. Progressive Residual Warmup for Language Model Pretraining 33 upvotes, #5 of 2026-03-09
  6. Reasoning Models Struggle to Control their Chains of Thought 28 upvotes, #6 of 2026-03-09
  7. RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies 24 upvotes, #7 of 2026-03-09
  8. Dynamic Chunking Diffusion Transformer 14 upvotes, #8 of 2026-03-09
  9. Physical Simulator In-the-Loop Video Generation 11 upvotes, #9 of 2026-03-09
  10. π-StepNFT: Wider Space Needs Finer Steps in Online RL for Flow-based VLAs 9 upvotes, #10 of 2026-03-09
  11. HiMAP-Travel: Hierarchical Multi-Agent Planning for Long-Horizon Constrained Travel 9 upvotes, #10 of 2026-03-09
  12. EffectMaker: Unifying Reasoning and Generation for Customized Visual Effect Creation 9 upvotes, #10 of 2026-03-09
  13. FlashPrefill: Instantaneous Pattern Discovery and Thresholding for Ultra-Fast Long-Context Prefilling 9 upvotes, #10 of 2026-03-09
  14. Mario: Multimodal Graph Reasoning with Large Language Models 8 upvotes, #14 of 2026-03-09
  15. Dynamic Model Routing and Cascading for Efficient LLM Inference: A Survey 4 upvotes, #15 of 2026-03-09
  16. DeepPresenter: Environment-Grounded Reflection for Agentic Presentation Generation 3 upvotes, #16 of 2026-03-09
  17. DreamCAD: Scaling Multi-modal CAD Generation using Differentiable Parametric Surfaces 3 upvotes, #16 of 2026-03-09
  18. WorldCache: Accelerating World Models for Free via Heterogeneous Token Caching 3 upvotes, #16 of 2026-03-09
  19. τ-Knowledge: Evaluating Conversational Agents over Unstructured Knowledge 2 upvotes, #19 of 2026-03-09
  20. PixARMesh: Autoregressive Mesh-Native Single-View Scene Reconstruction 2 upvotes, #19 of 2026-03-09
  21. Stabilizing Reinforcement Learning for Diffusion Language Models 2 upvotes, #19 of 2026-03-09
  22. Physics Informed Viscous Value Representations 1 upvotes, #22 of 2026-03-09
  23. Demystifying Action Space Design for Robotic Manipulation Policies 1 upvotes, #22 of 2026-03-09
  24. Operator Learning Using Weak Supervision from Walk-on-Spheres 1 upvotes, #22 of 2026-03-09
  25. Beyond the Grid: Layout-Informed Multi-Vector Retrieval with Parsed Visual Document Representations 1 upvotes, #22 of 2026-03-09
  26. IF-RewardBench: Benchmarking Judge Models for Instruction-Following Evaluation 1 upvotes, #22 of 2026-03-09
  27. nabla-Reasoner: LLM Reasoning via Test-Time Gradient Descent in Latent Space 1 upvotes, #22 of 2026-03-09
  28. Censored LLMs as a Natural Testbed for Secret Knowledge Elicitation 1 upvotes, #22 of 2026-03-09
  29. SLER-IR: Spherical Layer-wise Expert Routing for All-in-One Image Restoration 1 upvotes, #22 of 2026-03-09
  30. Layer by layer, module by module: Choose both for optimal OOD probing of ViT 0 upvotes, #30 of 2026-03-09
  31. Making Reconstruction FID Predictive of Diffusion Generation FID 1 upvotes, #30 of 2026-03-09

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.