Daily Papers of 2026-01-28

  1. AgentDoG: A Diagnostic Guardrail Framework for AI Agent Safety and Security 121 upvotes, #1 of 2026-01-28
  2. AdaReasoner: Dynamic Tool Orchestration for Iterative Visual Reasoning 47 upvotes, #2 of 2026-01-28
  3. A Pragmatic VLA Foundation Model 45 upvotes, #3 of 2026-01-28
  4. Youtu-VL: Unleashing Visual Potential via Unified Vision-Language Supervision 41 upvotes, #4 of 2026-01-28
  5. Visual Generation Unlocks Human-Like Reasoning through Multimodal World Models 25 upvotes, #5 of 2026-01-28
  6. Self-Distillation Enables Continual Learning 24 upvotes, #6 of 2026-01-28
  7. Post-LayerNorm Is Back: Stable, ExpressivE, and Deep 23 upvotes, #7 of 2026-01-28
  8. AVMeme Exam: A Multimodal Multilingual Multicultural Benchmark for LLMs' Contextual and Cultural Knowledge and Thinking 22 upvotes, #8 of 2026-01-28
  9. World Craft: Agentic Framework to Create Visualizable Worlds via Text 20 upvotes, #9 of 2026-01-28
  10. Towards Pixel-Level VLM Perception via Simple Points Prediction 16 upvotes, #10 of 2026-01-28
  11. FABLE: Forest-Based Adaptive Bi-Path LLM-Enhanced Retrieval for Multi-Document Reasoning 11 upvotes, #11 of 2026-01-28
  12. TriPlay-RL: Tri-Role Self-Play Reinforcement Learning for LLM Safety Alignment 10 upvotes, #12 of 2026-01-28
  13. HalluCitation Matters: Revealing the Impact of Hallucinated References with 300 Hallucinated Papers in ACL Conferences 7 upvotes, #13 of 2026-01-28
  14. Revisiting Parameter Server in LLM Post-Training 7 upvotes, #13 of 2026-01-28
  15. HyperAlign: Hypernetwork for Efficient Test-Time Alignment of Diffusion Models 6 upvotes, #15 of 2026-01-28
  16. Selective Steering: Norm-Preserving Control Through Discriminative Layer Selection 5 upvotes, #16 of 2026-01-28
  17. EvolVE: Evolutionary Search for LLM-based Verilog Generation and Optimization 4 upvotes, #17 of 2026-01-28
  18. DeFM: Learning Foundation Representations from Depth for Robotics 4 upvotes, #17 of 2026-01-28
  19. CooperBench: Why Coding Agents Cannot be Your Teammates Yet 3 upvotes, #19 of 2026-01-28
  20. GPCR-Filter: a deep learning framework for efficient and precise GPCR modulator discovery 2 upvotes, #20 of 2026-01-28
  21. Benchmarks Saturate When The Model Gets Smarter Than The Judge 2 upvotes, #20 of 2026-01-28

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.