Daily Papers of 2026-03-23

  1. HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoning 106 upvotes, #1 of 2026-03-23
  2. Astrolabe: Steering Forward-Process Reinforcement Learning for Distilled Autoregressive Video Models 106 upvotes, #1 of 2026-03-23
  3. TerraScope: Pixel-Grounded Visual Reasoning for Earth Observation 50 upvotes, #3 of 2026-03-23
  4. ProactiveBench: Benchmarking Proactiveness in Multimodal Large Language Models 41 upvotes, #4 of 2026-03-23
  5. Hyperagents 37 upvotes, #5 of 2026-03-23
  6. The Y-Combinator for LLMs: Solving Long-Context Rot with λ-Calculus 35 upvotes, #6 of 2026-03-23
  7. FlowScene: Style-Consistent Indoor Scene Generation with Multimodal Graph Rectified Flow 32 upvotes, #7 of 2026-03-23
  8. LumosX: Relate Any Identities with Their Attributes for Personalized Video Generation 22 upvotes, #8 of 2026-03-23
  9. Reasoning as Compression: Unifying Budget Forcing via the Conditional Information Bottleneck 21 upvotes, #9 of 2026-03-23
  10. A Subgoal-driven Framework for Improving Long-Horizon LLM Agents 19 upvotes, #10 of 2026-03-23
  11. Versatile Editing of Video Content, Actions, and Dynamics without Training 16 upvotes, #11 of 2026-03-23
  12. Deep Tabular Research via Continual Experience-Driven Execution 14 upvotes, #12 of 2026-03-23
  13. LoopRPT: Reinforcement Pre-Training for Looped Language Models 13 upvotes, #13 of 2026-03-23
  14. WorldAgents: Can Foundation Image Models be Agents for 3D World Models? 12 upvotes, #14 of 2026-03-23
  15. BEAVER: A Training-Free Hierarchical Prompt Compression Method via Structure-Aware Page Selection 11 upvotes, #15 of 2026-03-23
  16. HiMu: Hierarchical Multimodal Frame Selection for Long Video Question Answering 10 upvotes, #16 of 2026-03-23
  17. How Well Does Generative Recommendation Generalize? 10 upvotes, #16 of 2026-03-23
  18. Breaking the Capability Ceiling of LLM Post-Training by Reintroducing Markov States 9 upvotes, #18 of 2026-03-23
  19. EgoForge: Goal-Directed Egocentric World Simulator 9 upvotes, #18 of 2026-03-23
  20. Beyond Single Tokens: Distilling Discrete Diffusion Models via Discrete MMD 8 upvotes, #20 of 2026-03-23
  21. AgentDS Technical Report: Benchmarking the Future of Human-AI Collaboration in Domain-Specific Data Science 6 upvotes, #21 of 2026-03-23
  22. DROID-SLAM in the Wild 5 upvotes, #22 of 2026-03-23
  23. Do VLMs Need Vision Transformers? Evaluating State Space Models as Vision Encoders 5 upvotes, #22 of 2026-03-23
  24. Cooperation and Exploitation in LLM Policy Synthesis for Sequential Social Dilemmas 5 upvotes, #22 of 2026-03-23
  25. Teaching an Agent to Sketch One Part at a Time 5 upvotes, #22 of 2026-03-23
  26. Human-AI Synergy in Agentic Code Review 4 upvotes, #26 of 2026-03-23
  27. Language on Demand, Knowledge at Core: Composing LLMs with Encoder-Decoder Translation Models for Extensible Multilinguality 4 upvotes, #26 of 2026-03-23
  28. s2n-bignum-bench: A practical benchmark for evaluating low-level code reasoning of LLMs 3 upvotes, #28 of 2026-03-23
  29. Adaptive Layerwise Perturbation: Unifying Off-Policy Corrections for LLM RL 3 upvotes, #28 of 2026-03-23
  30. Probing Cultural Signals in Large Language Models through Author Profiling 2 upvotes, #30 of 2026-03-23
  31. TAPESTRY: From Geometry to Appearance via Consistent Turntable Videos 2 upvotes, #30 of 2026-03-23
  32. Multiscale Switch for Semi-Supervised and Contrastive Learning in Medical Ultrasound Image Segmentation 2 upvotes, #30 of 2026-03-23
  33. Automatic detection of Gen-AI texts: A comparative framework of neural models 2 upvotes, #30 of 2026-03-23
  34. CurveStream: Boosting Streaming Video Understanding in MLLMs via Curvature-Aware Hierarchical Visual Memory Management 2 upvotes, #30 of 2026-03-23
  35. ReLi3D: Relightable Multi-view 3D Reconstruction with Disentangled Illumination 2 upvotes, #30 of 2026-03-23
  36. ReLMXEL: Adaptive RL-Based Memory Controller with Explainable Energy and Latency Optimization 1 upvotes, #36 of 2026-03-23
  37. From Masks to Pixels and Meaning: A New Taxonomy, Benchmark, and Metrics for VLM Image Tampering 1 upvotes, #36 of 2026-03-23

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.