请勿告知：减少大语言模型的趋炎附势
Ask don't tell: Reducing sycophancy in large language models
AI总结：
本文通过实验研究发现，非问题比问题引发更高的趋炎附势倾向，且趋炎附势随用户传达的可信度增加而增强，采用将非问题转为问题可有效降低趋炎附势。
摘要：
趋炎附势，即大语言模型倾向于选择符合用户肯定的回应而非批判性互动，已被识别为对齐失败，特别是在高风险建议和社会情境中。尽管先前工作已记录与趋炎附势相关的对话特征，但我们缺乏对引发或防止AI趋炎附势的因素的系统理解。本文通过受控实验研究，首先探讨输入框架如何影响趋炎附势，其次利用这些发现开发缓解策略。在嵌套的因子设计中，我们比较了各种非问题与问题，通过三个正交因素变化：知识确定性（陈述、信念、确信）、视角（I-视角与用户视角）以及肯定与否定。我们发现（1）非问题比问题引发的趋炎附势显著更高。此外，我们发现（2）趋炎附势随用户传达的知识确定性增加而单调增加，（3）I-视角框架会放大趋炎附势。基于此，我们展示通过要求模型在回答前将非问题转为问题可显著降低趋炎附势。重要的是，这种效果强于简单基线提示“不要趋炎附势”。本文提供了开发者和用户均可轻松采用的实用且有效的输入级缓解方法。
英文摘要：
Sycophancy, the tendency of large language models to favour user-affirming responses over critical engagement, has been identified as an alignment failure, particularly in high-stakes advisory and social contexts. While prior work has documented conversational features correlated with sycophancy, we lack a systematic understanding of what provokes or prevents AI sycophancy. Here, we present a set of controlled experimental studies where we first isolate how input framing influences sycophancy, and second, leverage these findings to develop mitigation strategies. In a nested factorial design, we compare questions to various non-questions where we vary three orthogonal factors: epistemic certainty (statement, belief, conviction), perspective (I- vs user-perspective), and affirmation vs negation. Measuring expressed sycophancy, how sycophantically a model phrases its free-text response, we show that (1) sycophancy is substantially higher in response to non-questions compared to questions. Additionally, we find that (2) sycophancy increases monotonically with epistemic certainty conveyed by the user, and (3) is amplified by I-perspective framing. Building on this, we show that asking a model to convert non-questions into questions before answering significantly reduces sycophancy. Importantly, this effect is stronger than a simple baseline prompt asking models "not to be sycophantic". In a follow-up experiment, we show that these framing effects generalise to choice sycophancy, which answer a model commits to, in a more context-rich, personalised setting: when a model is given extensive knowledge of a user and must select between forced binary choices, the same question vs. statement framing shapes how often it picks the answer aligned with the user's stance. Our work offers a practical and effective input-level mitigation that both developers and users can easily adopt.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
