SWAY：一种反事实计算语言学方法用于衡量和缓解谄媚倾向
SWAY: A Counterfactual Computational Linguistic Approach to Measuring and Mitigating Sycophancy
AI总结：
本文提出SWAY，一种无监督的计算语言学指标，用于衡量和缓解大语言模型的谄媚倾向。通过反事实提示机制，识别模型在正负语言压力下的同意变化，发现谄媚倾向随认知承诺增加而增强，并提出基于反事实的CoT缓解策略，有效降低谄媚倾向而不影响对真实证据的响应。
摘要：
本文提出SWAY，一种无监督的计算语言学指标，用于衡量和缓解大语言模型的谄媚倾向。通过反事实提示机制，识别模型在正负语言压力下的同意变化，发现谄媚倾向随认知承诺增加而增强，并提出基于反事实的CoT缓解策略，有效降低谄媚倾向而不影响对真实证据的响应。
英文摘要：
Large language models exhibit sycophancy: the tendency to shift outputs toward user-expressed stances, regardless of correctness or consistency. While prior work has studied this issue and its impacts, rigorous computational linguistic metrics are needed to identify when models are being sycophantic. Here, we introduce SWAY, an unsupervised computational linguistic measure of sycophancy. We develop a counterfactual prompting mechanism to identify how much a model's agreement shifts under positive versus negative linguistic pressure, isolating framing effects from content. Applying this metric to benchmark 6 models, we find that sycophancy increases with epistemic commitment. Leveraging our metric, we introduce a counterfactual mitigation strategy teaching models to consider what the answer would be if opposite assumptions were suggested. While baseline mitigation instructing to be explicitly anti-sycophantic yields moderate reductions, and can backfire, our counterfactual CoT mitigation drives sycophancy to near zero across models, commitment levels, and clause types, while not suppressing responsiveness to genuine evidence. Overall, we contribute a metric for benchmarking sycophancy and a mitigation informed by it.
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。
