arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2601.13992cs.CLcs.AI
收藏

整体大于部分之和:一种兼容性感知的多教师CoT蒸馏框架

"The Whole Is Greater Than the Sum of Its Parts": A Compatibility-Aware Multi-Teacher CoT Distillation Framework

State Key Laboratory of Human-Machine Hybrid Augmented Intelligence, Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University(人机混合增强智能国家重点实验室,人工智能与机器人研究院,西安交通大学);Nankai University(南开大学);The Hong Kong University of Science and Technology(Guangzhou)(香港科技大学(广州));School of Software Engineering, State Key Laboratory of Human-Machine Hybrid Augmented Intelligence, Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University(软件学院,人机混合增强智能国家重点实验室,人工智能与机器人研究院,西安交通大学)

Jin Cui, Jiaqi Guo, Ruixuan Yang, Jiayi Lu, Jiepeng Zhou, Jiajun Xu, Jiangcheng Song, Boran Zhao, Pengju Ren

AI总结:

本文提出COMPACT框架,通过动态加权不同教师的梯度,结合多维指标提升学生模型的推理能力,有效整合多样化推理能力并减少灾难性遗忘。

摘要:

链式推理(CoT)推理赋予大语言模型(LLMs)显著能力,但通常需要极高的参数规模。CoT蒸馏作为一种有前景的范式,将推理能力转移到紧凑的学生模型(SLMs)中,但现有方法通常依赖单一教师,限制了学生潜力,因为个体LLMs常有不同能力偏倚且可能遭受灾难性遗忘。虽然利用多样教师似乎有吸引力,但有效融合其监督仍具挑战:教师-学生不兼容可能放大幻觉,被动监督无法确保真实逻辑内化。为此,我们引入COMPACT框架,通过动态加权教师梯度,基于多维指标评估学生实时兼容性:(1)基于图的共识过滤误导性推理路径;(2)基于互信息的适应性检测“顿悟时刻”以真正理解推理过程而非单纯模仿;(3)基于损失的难度评估学生对教师指导的接受度并防止负迁移。大量实验和潜在空间分析表明,COMPACT能有效整合多样化推理能力而不破坏模型原有知识结构,在各种基准测试中取得最佳性能并缓解灾难性遗忘。

英文摘要:

Chain-of-Thought (CoT) reasoning empowers Large Language Models (LLMs) with remarkable capabilities but typically requires prohibitive parameter scales. CoT distillation has emerged as a promising paradigm to transfer reasoning prowess into compact Student Models (SLMs), but existing approaches often rely on a solitary teacher, capping the student's potential since individual LLMs often exhibit distinct capability biases and may suffer from catastrophic forgetting. While leveraging diverse teachers seems appealing, effectively fusing their supervisions remains challenging: teacher-student incompatibility risks amplifying hallucinations, and passive supervision fails to ensure genuine logic internalization. To address this, we introduce COMPACT, a framework that adaptively fuses supervisions from different teachers by dynamically weighting teacher gradients based on the student's real-time compatibility evaluated by a multi-dimensional metric: (1) Graph-based Consensus to filter misleading rationales by identifying mainstream reasoning paths; (2) Mutual-Information-based Adaptability to detect "epiphany moments" for genuinely understanding the reasoning process rather than merely imitating; and (3) Loss-based Difficulty to assess student receptivity to the teacher's guidance and prevent negative transfer. Extensive experiments and latent space analysis demonstrate that COMPACT effectively integrates diverse reasoning capabilities without damaging the model's original knowledge structure, achieving state-of-the-art performance on various benchmarks while mitigating catastrophic forgetting.

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。

↑