Daily Papers of 2025-05-21

  1. Emerging Properties in Unified Multimodal Pretraining 124 upvotes, #1 of 2025-05-21
  2. SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training 58 upvotes, #2 of 2025-05-21
  3. Optimizing Anytime Reasoning via Budget Relative Policy Optimization 34 upvotes, #3 of 2025-05-21
  4. Neurosymbolic Diffusion Models 33 upvotes, #4 of 2025-05-21
  5. Reward Reasoning Model 32 upvotes, #5 of 2025-05-21
  6. Visual Agentic Reinforcement Fine-Tuning 31 upvotes, #6 of 2025-05-21
  7. VisualQuality-R1: Reasoning-Induced Image Quality Assessment via Reinforcement Learning to Rank 30 upvotes, #7 of 2025-05-21
  8. Latent Flow Transformer 27 upvotes, #8 of 2025-05-21
  9. The Aloe Family Recipe for Open and Specialized Healthcare LLMs 26 upvotes, #9 of 2025-05-21
  10. General-Reasoner: Advancing LLM Reasoning Across All Domains 20 upvotes, #10 of 2025-05-21
  11. Reasoning Models Better Express Their Confidence 18 upvotes, #11 of 2025-05-21
  12. Think Only When You Need with Large Hybrid-Reasoning Models 18 upvotes, #11 of 2025-05-21
  13. Reasoning Path Compression: Compressing Generation Trajectories for Efficient LLM Reasoning 16 upvotes, #13 of 2025-05-21
  14. Visionary-R1: Mitigating Shortcuts in Visual Reasoning with Reinforcement Learning 15 upvotes, #14 of 2025-05-21
  15. Hunyuan-Game: Industrial-grade Intelligent Game Creation Model 14 upvotes, #15 of 2025-05-21
  16. VideoEval-Pro: Robust and Realistic Long Video Understanding Evaluation 14 upvotes, #15 of 2025-05-21
  17. Exploring Federated Pruning for Large Language Models 13 upvotes, #17 of 2025-05-21
  18. Training-Free Watermarking for Autoregressive Image Generation 12 upvotes, #18 of 2025-05-21
  19. CS-Sum: A Benchmark for Code-Switching Dialogue Summarization and the Limits of Large Language Models 11 upvotes, #19 of 2025-05-21
  20. SSR: Enhancing Depth Perception in Vision-Language Models via Rationale-Guided Spatial Reasoning 10 upvotes, #20 of 2025-05-21
  21. Fine-tuning Quantized Neural Networks with Zeroth-order Optimization 10 upvotes, #20 of 2025-05-21
  22. Visual Instruction Bottleneck Tuning 10 upvotes, #20 of 2025-05-21
  23. Towards eliciting latent knowledge from LLMs with mechanistic interpretability 9 upvotes, #23 of 2025-05-21
  24. NExT-Search: Rebuilding User Feedback Ecosystem for Generative AI Search 9 upvotes, #23 of 2025-05-21
  25. Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training 9 upvotes, #23 of 2025-05-21
  26. The Hallucination Tax of Reinforcement Finetuning 8 upvotes, #26 of 2025-05-21
  27. Not All Correct Answers Are Equal: Why Your Distillation Source Matters 8 upvotes, #26 of 2025-05-21
  28. Lessons from Defending Gemini Against Indirect Prompt Injections 8 upvotes, #26 of 2025-05-21
  29. Vox-Profile: A Speech Foundation Model Benchmark for Characterizing Diverse Speaker and Speech Traits 8 upvotes, #26 of 2025-05-21
  30. Truth Neurons 7 upvotes, #30 of 2025-05-21
  31. Warm Up Before You Train: Unlocking General Reasoning in Resource-Constrained Settings 7 upvotes, #30 of 2025-05-21
  32. MIGRATION-BENCH: Repository-Level Code Migration Benchmark from Java 8 6 upvotes, #32 of 2025-05-21
  33. Phare: A Safety Probe for Large Language Models 6 upvotes, #32 of 2025-05-21
  34. Fixing 7,400 Bugs for 1$: Cheap Crash-Site Program Repair 6 upvotes, #32 of 2025-05-21
  35. Rethinking Optimal Verification Granularity for Compute-Efficient Test-Time Scaling 5 upvotes, #35 of 2025-05-21
  36. Solve-Detect-Verify: Inference-Time Scaling with Flexible Generative Verifier 5 upvotes, #35 of 2025-05-21
  37. CompeteSMoE -- Statistically Guaranteed Mixture of Experts Training via Competition 5 upvotes, #35 of 2025-05-21
  38. Bidirectional LMs are Better Knowledge Memorizers? A Benchmark for Real-world Knowledge Injection 4 upvotes, #38 of 2025-05-21
  39. CoIn: Counting the Invisible Reasoning Tokens in Commercial Opaque LLM APIs 4 upvotes, #38 of 2025-05-21
  40. Incorporating brain-inspired mechanisms for multimodal learning in artificial intelligence 3 upvotes, #40 of 2025-05-21
  41. To Bias or Not to Bias: Detecting bias in News with bias-detector 3 upvotes, #40 of 2025-05-21
  42. Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas 3 upvotes, #40 of 2025-05-21
  43. Understanding Gen Alpha Digital Language: Evaluation of LLM Safety Systems for Content Moderation 2 upvotes, #43 of 2025-05-21
  44. Masking in Multi-hop QA: An Analysis of How Language Models Perform with Context Permutation 2 upvotes, #43 of 2025-05-21
  45. Learning to Highlight Audio by Watching Movies 2 upvotes, #43 of 2025-05-21
  46. GeoRanker: Distance-Aware Ranking for Worldwide Image Geolocalization 2 upvotes, #43 of 2025-05-21
  47. Tokenization Constraints in LLMs: A Study of Symbolic and Arithmetic Reasoning Limits 2 upvotes, #43 of 2025-05-21
  48. The Distracting Effect: Understanding Irrelevant Passages in RAG 1 upvotes, #48 of 2025-05-21
  49. Void in Language Models 1 upvotes, #48 of 2025-05-21
  50. Dynadiff: Single-stage Decoding of Images from Continuously Evolving fMRI 1 upvotes, #48 of 2025-05-21
  51. KERL: Knowledge-Enhanced Personalized Recipe Recommendation using Large Language Models 1 upvotes, #48 of 2025-05-21
  52. Object-Centric Representations Improve Policy Generalization in Robot Manipulation 0 upvotes, #52 of 2025-05-21
  53. Towards Embodied Cognition in Robots via Spatially Grounded Synthetic Worlds 0 upvotes, #52 of 2025-05-21

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.