一作ACL 2026Poster· Natural Language Processing
The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysis
Zihao Wei · Liang Pang · Jiahao Liu · Wenjie Shi · Jingcheng Deng · Shicheng Xu · Zenghao Duan · Jingang Wang · Fei Sun · Huawei Shen · Xueqi Cheng
摘要
Test-time scaling via explicit reasoning trajectories significantly boosts large language model (LLM) performance but often triggers overthinking. To explore this, we analyze reasoning through two lenses: Reasoning Length Dynamics, which reveals a compensatory trade-off between thinking and answe...