arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.05355cs.AIcs.CL

ETR:熵趋势奖励用于高效推理链推理

ETR: Entropy Trend Reward for Efficient Chain-of-Thought Reasoning

University of Toronto(多伦多大学);McMaster University(麦克马斯特大学);Concordia University(康考迪亚大学);University of Ottawa(渥太华大学)

Xuan Xiong, Huan Liu, Li Gu, Zhixiang Chi, Yue Qiu, Yuanhao Yu, Yang Wang

AI总结:

本文提出ETR,通过轨迹感知的目标促进逐步不确定性降低,提升推理效率与准确性,实验表明在多个基准上显著提升性能。

摘要:

本文提出ETR,通过轨迹感知的目标促进逐步不确定性降低,提升推理效率与准确性,实验表明在多个基准上显著提升性能。

英文摘要:

Chain-of-thought (CoT) reasoning improves large language model performance on complex tasks, but often produces excessively long and inefficient reasoning traces. Existing methods shorten CoTs using length penalties or global entropy reduction, implicitly assuming that low uncertainty is desirable throughout reasoning. We show instead that reasoning efficiency is governed by the trajectory of uncertainty. CoTs with dominant downward entropy trends are substantially shorter. Motivated by this insight, we propose Entropy Trend Reward (ETR), a trajectory-aware objective that encourages progressive uncertainty reduction while allowing limited local exploration. We integrate ETR into Group Relative Policy Optimization (GRPO) and evaluate it across multiple reasoning models and challenging benchmarks. ETR consistently achieves a superior accuracy-efficiency tradeoff, improving DeepSeek-R1-Distill-7B by 9.9% in accuracy while reducing CoT length by 67% across four benchmarks. Code is available at https://github.com/Xuan1030/ETR

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑