质疑问题本身：维持推理模型的自进化
原标题：Paper page - Questioning the Questions: Sustaining Self-Evolution in Reasoning Models
AI 摘要
该论文研究自进化推理模型在反复自我训练中出现的性能崩溃问题，指出自生成问题存在无效问题增多和数学等价问题重复两类质量缺陷，且基于词汇相似度的多样性控制无法识别数学等价问题。作者提出 R-Quest，利用问题有效性与新颖性反馈引导自进化：先训练求解器识别并拒绝无效问题，再用其判断引导提问者奖励并过滤训练数据，同时用冻结基座模型比较问题对以提供新颖性反馈。实验显示 R-Quest 在两个模型家族的 12 个数学推理、通用推理和代码生成基准上取得最高平均性能，并在十轮自进化中保持稳定增益，最终轮超过 R-Zero 17.32 分。
正文节选
Questioning the Questions: Sustaining Self-Evolution in Reasoning Models Abstract Self-evolving reasoning models learn from their own generated questions, yet repeated self-training can lead to performance collapse. In this paper, we investigate why performance deteriorates over successive rounds and how to sustain self-evolution. Our analysis identifies two recurring quality problems in self-generated questions: invalid questions and repeated variants of the same mathematical questions. First,