啄木鸟蒸馏：弱模型诊断强模型推理错误
原标题：Woodpecker Distillation: Weak Models Diagnose Reasoning Bugs in Strong Models
AI 摘要
arXiv 上发布了一篇题为“Woodpecker Distillation: Weak Models Diagnose Reasoning Bugs in Strong Models”的论文。论文指出，大型语言模型在推理任务中的失败往往源于中间步骤的局部推理错误，而非整体能力不足。作者提出了一种名为 Woodpecker Distillation 的弱到强训练框架，通过对比成功与不成功的弱模型补丁，构建纠正性的教师分布，并将该信号蒸馏到强模型中。在数学推理基准上的实验表明，该方法能持续提升强模型性能，并优于直接模仿基线。
正文节选
Computer Science > Artificial Intelligence Title:Woodpecker Distillation: Weak Models Diagnose Reasoning Bugs in Strong Models View PDF HTML (experimental) Abstract:Large language models often fail on reasoning tasks despite possessing the capability to solve them. We argue that many such failures arise from localized reasoning bugs in intermediate steps rather than from global incompetence. We show that these bugs are frequently repairable: inserting a short patch generated by a wea