Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning Paper • 2506.01939 • Published 3 days ago • 120
Emerging Properties in Unified Multimodal Pretraining Paper • 2505.14683 • Published 16 days ago • 129
AttentionInfluence: Adopting Attention Head Influence for Weak-to-Strong Pretraining Data Selection Paper • 2505.07293 • Published 24 days ago • 26
Unified Multimodal Understanding and Generation Models: Advances, Challenges, and Opportunities Paper • 2505.02567 • Published May 5 • 74
COIG-P: A High-Quality and Large-Scale Chinese Preference Dataset for Alignment with Human Values Paper • 2504.05535 • Published Apr 7 • 44