Daily Papers of 2026-03-19

  1. Efficient Reasoning with Balanced Thinking 141 upvotes, #1 of 2026-03-19
  2. MetaClaw: Just Talk -- An Agent That Meta-Learns and Evolves in the Wild 128 upvotes, #2 of 2026-03-19
  3. Video-CoE: Reinforcing Video Event Prediction via Chain of Events 90 upvotes, #3 of 2026-03-19
  4. MosaicMem: Hybrid Spatial Memory for Controllable Video World Models 85 upvotes, #4 of 2026-03-19
  5. Alignment Makes Language Models Normative, Not Descriptive 46 upvotes, #5 of 2026-03-19
  6. Complementary Reinforcement Learning 35 upvotes, #6 of 2026-03-19
  7. LoST: Level of Semantics Tokenization for 3D Shapes 30 upvotes, #7 of 2026-03-19
  8. When AI Navigates the Fog of War 28 upvotes, #8 of 2026-03-19
  9. GigaWorld-Policy: An Efficient Action-Centered World--Action Model 24 upvotes, #9 of 2026-03-19
  10. V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning 22 upvotes, #10 of 2026-03-19
  11. BenchPreS: A Benchmark for Context-Aware Personalized Preference Selectivity of Persistent-Memory LLMs 22 upvotes, #10 of 2026-03-19
  12. Look Before Acting: Enhancing Vision Foundation Representations for Vision-Language-Action Models 21 upvotes, #12 of 2026-03-19
  13. Temporal Gains, Spatial Costs: Revisiting Video Fine-Tuning in Multimodal Large Language Models 20 upvotes, #13 of 2026-03-19
  14. ESPIRE: A Diagnostic Benchmark for Embodied Spatial Reasoning of Vision-Language Models 13 upvotes, #14 of 2026-03-19
  15. AdaMem: Adaptive User-Centric Memory for Long-Horizon Dialogue Agents 13 upvotes, #14 of 2026-03-19
  16. Efficient Exploration at Scale 12 upvotes, #16 of 2026-03-19
  17. Unified Spatio-Temporal Token Scoring for Efficient Video VLMs 12 upvotes, #16 of 2026-03-19
  18. Conservative Offline Robot Policy Learning via Posterior-Transition Reweighting 11 upvotes, #18 of 2026-03-19
  19. Stereo World Model: Camera-Guided Stereo Video Generation 11 upvotes, #18 of 2026-03-19
  20. Expert Threshold Routing for Autoregressive Language Modeling with Dynamic Computation Allocation and Load Balancing 10 upvotes, #20 of 2026-03-19
  21. RAMP: Reinforcement Adaptive Mixed Precision Quantization for Efficient On Device LLM Inference 7 upvotes, #21 of 2026-03-19
  22. Efficient Training-Free Multi-Token Prediction via Embedding-Space Probing 7 upvotes, #21 of 2026-03-19
  23. LaDe: Unified Multi-Layered Graphic Media Generation and Decomposition 6 upvotes, #23 of 2026-03-19
  24. ACE-LoRA: Graph-Attentive Context Enhancement for Parameter-Efficient Adaptation of Medical Vision-Language Models 4 upvotes, #24 of 2026-03-19
  25. From Prior to Pro: Efficient Skill Mastery via Distribution Contractive RL Finetuning 3 upvotes, #25 of 2026-03-19
  26. AI Scientist via Synthetic Task Scaling 3 upvotes, #25 of 2026-03-19
  27. FINER: MLLMs Hallucinate under Fine-grained Negative Queries 3 upvotes, #25 of 2026-03-19
  28. VideoAtlas: Navigating Long-Form Video in Logarithmic Compute 3 upvotes, #25 of 2026-03-19
  29. Coherent Human-Scene Reconstruction from Multi-Person Multi-View Video in a Single Pass 2 upvotes, #29 of 2026-03-19
  30. AdapterTune: Zero-Initialized Low-Rank Adapters for Frozen Vision Transformers 1 upvotes, #30 of 2026-03-19
  31. HeBA: Heterogeneous Bottleneck Adapters for Robust Vision-Language Models 1 upvotes, #30 of 2026-03-19
  32. PRISM: Demystifying Retention and Interaction in Mid-Training 1 upvotes, #30 of 2026-03-19
  33. Fanar-Sadiq: A Multi-Agent Architecture for Grounded Islamic QA 0 upvotes, #33 of 2026-03-19

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.