Daily Papers of 2026-04-07

  1. Adam's Law: Textual Frequency Law on Large Language Models 468 upvotes, #1 of 2026-04-07
  2. OpenWorldLib: A Unified Codebase and Definition of Advanced World Models 200 upvotes, #2 of 2026-04-07
  3. MinerU2.5-Pro: Pushing the Limits of Data-Centric Document Parsing at Scale 117 upvotes, #3 of 2026-04-07
  4. TriAttention: Efficient Long Reasoning with Trigonometric KV Compression 106 upvotes, #4 of 2026-04-07
  5. LIBERO-Para: A Diagnostic Benchmark and Metrics for Paraphrase Robustness in VLA Models 81 upvotes, #5 of 2026-04-07
  6. Memory Intelligence Agent 56 upvotes, #6 of 2026-04-07
  7. AURA: Always-On Understanding and Real-Time Assistance via Video Streams 50 upvotes, #7 of 2026-04-07
  8. Can LLMs Learn to Reason Robustly under Noisy Supervision? 42 upvotes, #8 of 2026-04-07
  9. FileGram: Grounding Agent Personalization in File-System Behavioral Traces 40 upvotes, #9 of 2026-04-07
  10. ClawArena: Benchmarking AI Agents in Evolving Information Environments 36 upvotes, #10 of 2026-04-07
  11. SpatialEdit: Benchmarking Fine-Grained Image Spatial Editing 35 upvotes, #11 of 2026-04-07
  12. BidirLM: From Text to Omnimodal Bidirectional Encoders by Adapting and Composing Causal LLMs 33 upvotes, #12 of 2026-04-07
  13. Self-Execution Simulation Improves Coding Models 33 upvotes, #12 of 2026-04-07
  14. LightThinker++: From Reasoning Compression to Memory Management 33 upvotes, #12 of 2026-04-07
  15. SkillX: Automatically Constructing Skill Knowledge Bases for Agents 31 upvotes, #15 of 2026-04-07
  16. Unifying Group-Relative and Self-Distillation Policy Optimization via Sample Routing 30 upvotes, #16 of 2026-04-07
  17. Vero: An Open RL Recipe for General Visual Reasoning 30 upvotes, #16 of 2026-04-07
  18. Your Agent, Their Asset: A Real-World Safety Analysis of OpenClaw 22 upvotes, #18 of 2026-04-07
  19. Learning to Learn-at-Test-Time: Language Agents with Learnable Adaptation Policies 15 upvotes, #19 of 2026-04-07
  20. PLUME: Latent Reasoning Based Universal Multimodal Embedding 15 upvotes, #19 of 2026-04-07
  21. Mimic Intent, Not Just Trajectories 14 upvotes, #21 of 2026-04-07
  22. Synthetic Sandbox for Training Machine Learning Engineering Agents 14 upvotes, #21 of 2026-04-07
  23. ONE-SHOT: Compositional Human-Environment Video Synthesis via Spatial-Decoupled Motion Injection and Hybrid Context Integration 13 upvotes, #23 of 2026-04-07
  24. Less Detail, Better Answers: Degradation-Driven Prompting for VQA 13 upvotes, #23 of 2026-04-07
  25. Paper Espresso: From Paper Overload to Research Insight 12 upvotes, #25 of 2026-04-07
  26. AvatarPointillist: AutoRegressive 4D Gaussian Avatarization 12 upvotes, #25 of 2026-04-07
  27. Scaling Teams or Scaling Time? Memory Enabled Lifelong Learning in LLM Multi-Agent Systems 10 upvotes, #27 of 2026-04-07
  28. The Geometric Alignment Tax: Tokenization vs. Continuous Geometry in Scientific Foundation Models 10 upvotes, #27 of 2026-04-07
  29. HDP: A Lightweight Cryptographic Protocol for Human Delegation Provenance in Agentic AI Systems 10 upvotes, #27 of 2026-04-07
  30. CLEAR: Unlocking Generative Potential for Degraded Image Understanding in Unified Multimodal Models 10 upvotes, #27 of 2026-04-07
  31. SciLT: Long-Tailed Classification in Scientific Image Domains 8 upvotes, #31 of 2026-04-07
  32. Type-Checked Compliance: Deterministic Guardrails for Agentic Financial Systems Using Lean 4 Theorem Proving 7 upvotes, #32 of 2026-04-07
  33. Do Audio-Visual Large Language Models Really See and Hear? 7 upvotes, #32 of 2026-04-07
  34. Emergent Compositional Communication for Latent World Properties 7 upvotes, #32 of 2026-04-07
  35. Cog-DRIFT: Exploration on Adaptively Reformulated Instances Enables Learning from Hard Reasoning Problems 7 upvotes, #32 of 2026-04-07
  36. Locally Confident, Globally Stuck: The Quality-Exploration Dilemma in Diffusion Language Models 5 upvotes, #36 of 2026-04-07
  37. POEMetric: The Last Stanza of Humanity 4 upvotes, #37 of 2026-04-07

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.