Daily Papers of 2026-04-15

  1. KnowRL: Boosting LLM Reasoning via Reinforcement Learning with Minimal-Sufficient Knowledge Guidance 98 upvotes, #1 of 2026-04-15
  2. Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe 85 upvotes, #2 of 2026-04-15
  3. Lyra 2.0: Explorable Generative 3D Worlds 37 upvotes, #3 of 2026-04-15
  4. Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning 36 upvotes, #4 of 2026-04-15
  5. Toward Autonomous Long-Horizon Engineering for ML Research 34 upvotes, #5 of 2026-04-15
  6. Turing Test on Screen: A Benchmark for Mobile GUI Agent Humanization 30 upvotes, #6 of 2026-04-15
  7. SPPO: Sequence-Level PPO for Long-Horizon Reasoning Tasks 29 upvotes, #7 of 2026-04-15
  8. BERT-as-a-Judge: A Robust Alternative to Lexical Methods for Efficient Reference-Based LLM Evaluation 29 upvotes, #7 of 2026-04-15
  9. The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents 24 upvotes, #9 of 2026-04-15
  10. LARY: A Latent Action Representation Yielding Benchmark for Generalizable Vision-to-Action Alignment 20 upvotes, #10 of 2026-04-15
  11. Towards Long-horizon Agentic Multimodal Search 20 upvotes, #10 of 2026-04-15
  12. Hierarchical SVG Tokenization: Learning Compact Visual Programs for Scalable Vector Graphics Modeling 18 upvotes, #12 of 2026-04-15
  13. Many-Tier Instruction Hierarchy in LLM Agents 16 upvotes, #13 of 2026-04-15
  14. Rethinking the Diffusion Model from a Langevin Perspective 15 upvotes, #14 of 2026-04-15
  15. Generative Refinement Networks for Visual Synthesis 15 upvotes, #14 of 2026-04-15
  16. Habitat-GS: A High-Fidelity Navigation Simulator with Dynamic Gaussian Splatting 14 upvotes, #16 of 2026-04-15
  17. Feed-Forward 3D Scene Modeling: A Problem-Driven Perspective 14 upvotes, #16 of 2026-04-15
  18. Self-Adversarial One Step Generation via Condition Shifting 13 upvotes, #18 of 2026-04-15
  19. Lightning OPD: Efficient Post-Training for Large Reasoning Models with Offline On-Policy Distillation 12 upvotes, #19 of 2026-04-15
  20. You Only Judge Once: Multi-response Reward Modeling in a Single Forward Pass 11 upvotes, #20 of 2026-04-15
  21. Masked by Consensus: Disentangling Privileged Knowledge in LLM Correctness 9 upvotes, #21 of 2026-04-15
  22. Grid2Matrix: Revealing Digital Agnosia in Vision-Language Models 8 upvotes, #22 of 2026-04-15
  23. Do Thought Streams Matter? Evaluating Reasoning in Gemini Vision-Language Models for Video Scene Understanding 7 upvotes, #23 of 2026-04-15
  24. Parcae: Scaling Laws For Stable Looped Language Models 6 upvotes, #24 of 2026-04-15
  25. Accelerating Speculative Decoding with Block Diffusion Draft Trees 6 upvotes, #24 of 2026-04-15
  26. LASA: Language-Agnostic Semantic Alignment at the Semantic Bottleneck for LLM Safety 5 upvotes, #26 of 2026-04-15
  27. GlotOCR Bench: OCR Models Still Struggle Beyond a Handful of Unicode Scripts 5 upvotes, #26 of 2026-04-15
  28. Spec Kit Agents: Context-Grounded Agentic Workflows 4 upvotes, #28 of 2026-04-15
  29. VideoFlexTok: Flexible-Length Coarse-to-Fine Video Tokenization 4 upvotes, #28 of 2026-04-15
  30. PokeRL: Reinforcement Learning for Pokemon Red 3 upvotes, #30 of 2026-04-15
  31. Seeing Through Touch: Tactile-Driven Visual Localization of Material Regions 3 upvotes, #30 of 2026-04-15
  32. Domain-Specific Latent Representations Improve the Fidelity of Diffusion-Based Medical Image Super-Resolution 3 upvotes, #30 of 2026-04-15
  33. Learning Versatile Humanoid Manipulation with Touch Dreaming 3 upvotes, #30 of 2026-04-15
  34. Spatial Competence Benchmark 2 upvotes, #34 of 2026-04-15
  35. When Reasoning Models Hurt Behavioral Simulation: A Solver-Sampler Mismatch in Multi-Agent LLM Negotiation 2 upvotes, #34 of 2026-04-15
  36. Beyond Perception Errors: Semantic Fixation in Large Vision-Language Models 2 upvotes, #34 of 2026-04-15
  37. CONSCIENTIA: Can LLM Agents Learn to Strategize? Emergent Deception and Trust in a Multi-Agent NYC Simulation 1 upvotes, #37 of 2026-04-15
  38. 3DTV: A Feedforward Interpolation Network for Real-Time View Synthesis 1 upvotes, #37 of 2026-04-15
  39. SpotSound: Enhancing Large Audio-Language Models with Fine-Grained Temporal Grounding 1 upvotes, #37 of 2026-04-15

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.