Junyang Lin

8 papers Alibaba Group

• Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning
• CARE: Decoding-Time Safety Alignment via Rollback and Introspection Intervention
• Chain of Execution Supervision Promotes General Reasoning in Large Language Models
• Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free
• Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free
• Parallel Scaling Law for Language Models
• PolyMath: Evaluating Mathematical Reasoning in Multilingual Contexts
• Teaching Language Models to Reason with Tools