[论文] Beyond the Answer: Student-Conditioned Trace Construction for Reasonin...

论文概要

研究领域: NLP
作者: Shaoang Li, Yanhang Shi, Yufei Li, Mingfu Liang, Frank Shyu, Sandeep Pandey, Luke Simon, Xi Liu, Jian Li
发布时间: 2026-10-16
arXiv: 2610.23831

中文摘要

一条推理痕迹可以伴随着正确答案，却无法为特定学生提供有效的监督。我们研究如何构建推理痕迹以提升超越仅答案监督的学习效果，同时考虑其中间内容和对目标学习者的可及性。我们提出 ReDiST——一个面向证据丰富推理的学生条件化痕迹构建与蒸馏框架。从学生自己的尝试出发，ReDiST 对照给定证据修正答案正确的草稿，并利用已验证答案的额外指导修复不成功的草稿。一个冻结的、与学生匹配的阅读器根据问题、证据和候选痕迹，使用答案恢复和已验证答案概率来筛选和排序原始及编辑后的候选。我们在蒸馏选定的目标时，通过辅助答案损失平衡痕迹监督和最终答案监督。构建和选择离线进行，推理仅使用蒸馏后的学生。问答基准上的实验表明，ReDiST 能提升源任务和跨基准性能，优于仅答案、原始痕迹以及所评估的痕迹选择和重写基线。在不同的监督方案中，性能排名在源基准和迁移基准之间可能不同，突出了评估跨分布的痕迹教学效用的重要性。

原文摘要

A reasoning trace can accompany a correct answer without providing effective supervision for a particular student. We study how to construct traces that improve learning beyond answer-only supervision by considering both their intermediate content and their accessibility to the target learner. We introduce ReDiST, a student-conditioned trace-construction and distillation framework for evidence-rich reasoning. Starting from the student's own attempts, ReDiST revises answer-correct drafts against the supplied evidence and repairs unsuccessful drafts with additional guidance from the verified answer. A frozen student-matched reader screens and ranks original and edited candidates using answer recovery and verified-answer probability, given the question, evidence, and candidate trace. We disti...

自动采集于 2026-10-20

#论文 #arXiv #NLP #小凯