揭示大模型讨好行为研究的方法困境，强调需引入人类评估
Sycophancy Claims about Language Models: The Missing Human-in-the-Loop
- 梳理五种衡量大模型讨好行为的操作化定义
- 指出现有研究忽视人类感知，难以区分讨好与对齐概念
- 建议未来研究加入人类在环评估，提升可靠性
大语言模型（LLMs）的讨好响应模式在文献中被越来越多地提及。本文回顾了测量LLM讨好行为的方法论挑战，识别出五个核心操作化定义。尽管讨好行为本质上是人为导向的，但当前研究未评估人类感知。分析表明，区分讨好响应与人工智能对齐中的相关概念存在困难，并为未来研究提供了可操作的改进建议。
原文摘要 · Abstract (English)
Sycophantic response patterns in Large Language Models (LLMs) have been increasingly claimed in the literature. We review methodological challenges in measuring LLM sycophancy and identify five core operationalizations. Despite sycophancy being inherently human-centric, current research does not evaluate human perception. Our analysis highlights the difficulties in distinguishing sycophantic responses from related concepts in AI alignment and offers actionable recommendations for future research.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。