Self-Correction Bench: Uncovering and Addressing the Self-Correction Blind Spot in Large Language Models
Self-Correction Bench:揭示并解决大语言模型中的自校正盲点
机构 * Independent Researcher(独立研究者)
专题命中 复杂问题求解 :self-correction(title,title_cn);reasoning(abstract);分类 cs.CL、cs.AI、cs.LG
AI总结 该研究提出Self-Correction Bench框架,发现大语言模型存在64.5%的自校正盲点,经微调或添加“Wait”可显著降低该盲点,揭示了自校正能力未激活的机制。
Comments Accepted to COLM 2026