Daily Papers of 2026-04-23

  1. LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language Model 232 upvotes, #1 of 2026-04-23
  2. Near-Future Policy Optimization 68 upvotes, #2 of 2026-04-23
  3. DR-Venus: Towards Frontier Edge-Scale Deep Research Agents with Only 10K Open Data 49 upvotes, #3 of 2026-04-23
  4. Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges 29 upvotes, #4 of 2026-04-23
  5. OpenMobile: Building Open Mobile Agents with Task and Trajectory Synthesis 27 upvotes, #5 of 2026-04-23
  6. DeVI: Physics-based Dexterous Human-Object Interaction via Synthetic Video Imitation 24 upvotes, #6 of 2026-04-23
  7. Exploring Spatial Intelligence from a Generative Perspective 21 upvotes, #7 of 2026-04-23
  8. A Self-Evolving Framework for Efficient Terminal Agents via Observational Context Compression 18 upvotes, #8 of 2026-04-23
  9. Expert Upcycling: Shifting the Compute-Efficient Frontier of Mixture-of-Experts 17 upvotes, #9 of 2026-04-23
  10. C-GenReg: Training-Free 3D Point Cloud Registration by Multi-View-Consistent Geometry-to-Image Generation with Probabilistic Modalities Fusion 15 upvotes, #10 of 2026-04-23
  11. Image Generators are Generalist Vision Learners 14 upvotes, #11 of 2026-04-23
  12. SWE-chat: Coding Agent Interactions From Real Users in the Wild 12 upvotes, #12 of 2026-04-23
  13. WavAlign: Enhancing Intelligence and Expressiveness in Spoken Dialogue Models via Adaptive Hybrid Post-Training 11 upvotes, #13 of 2026-04-23
  14. Scaling Test-Time Compute for Agentic Coding 10 upvotes, #14 of 2026-04-23
  15. Abstain-R1: Calibrated Abstention and Post-Refusal Clarification via Verifiable RL 9 upvotes, #15 of 2026-04-23
  16. Self-Evolving LLM Memory Extraction Across Heterogeneous Tasks 7 upvotes, #16 of 2026-04-23
  17. Convergent Evolution: How Different Language Models Learn Similar Number Representations 7 upvotes, #16 of 2026-04-23
  18. Cortex 2.0: Grounding World Models in Real-World Industrial Deployment 6 upvotes, #18 of 2026-04-23
  19. Tadabur: A Large-Scale Quran Audio Dataset 5 upvotes, #19 of 2026-04-23
  20. Chasing the Public Score: User Pressure and Evaluation Exploitation in Coding Agent Workflows 5 upvotes, #19 of 2026-04-23
  21. AI scientists produce results without reasoning scientifically 4 upvotes, #21 of 2026-04-23
  22. SAVOIR: Learning Social Savoir-Faire via Shapley-based Reward Attribution 4 upvotes, #21 of 2026-04-23
  23. Visual Reasoning through Tool-supervised Reinforcement Learning 4 upvotes, #21 of 2026-04-23
  24. Diverse Dictionary Learning 3 upvotes, #24 of 2026-04-23
  25. ReImagine: Rethinking Controllable High-Quality Human Video Generation via Image-First Synthesis 3 upvotes, #24 of 2026-04-23
  26. Streaming Structured Inference with Flash-SemiCRF 2 upvotes, #26 of 2026-04-23
  27. MMCORE: MultiModal COnnection with Representation Aligned Latent Embeddings 2 upvotes, #26 of 2026-04-23
  28. CreativeGame:Toward Mechanic-Aware Creative Game Generation 2 upvotes, #26 of 2026-04-23
  29. COMPASS: COntinual Multilingual PEFT with Adaptive Semantic Sampling 2 upvotes, #26 of 2026-04-23
  30. Benign Fine-Tuning Breaks Safety Alignment in Audio LLMs 1 upvotes, #30 of 2026-04-23

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.