作者位次 #3/8ACL 2026Poster· Natural Language Processing
Not All Tokens Matter: Towards Efficient LLM Reasoning via Token Significance in Reinforcement Learning
Hanbing Liu · Lang Cao · Yuanyi Ren · Mengyu Zhou · Haoyu Dong · Xiaojun Ma · Shi Han · Dongmei Zhang
摘要
Large language models (LLMs) show strong reasoning abilities but often produce unnecessarily long explanations that reduce efficiency. Although reinforcement learning (RL) has been used to improve reasoning, most methods focus on accuracy and rely on uniform length-based rewards that overlook the...