EntroCut: Entropy-Guided Adaptive Truncation for Efficient Chain-of-Thought Reasoning in Small-scale Large Reasoning Models
EntroCut: 基于熵的自适应截断以提高小型大推理模型中的链式推理效率
机构 * 1 Beihang University, School of Computer Science \& Engineering, China 2 Zhongguancun Laboratory, China 3 Hangzhou Innovation Institute, China
专题命中 复杂问题求解 :reasoning(title,abstract);chain-of-thought(title,abstract);分类 cs.AI
AI总结 EntroCut通过基于熵的自适应截断方法,显著提高小型大推理模型的推理效率,同时保持较高的准确性。
Comments Accepted by ICASSP26