压力下的谄媚：在科学问答中通过对抗性对话评估与缓解谄媚偏差
Sycophancy under Pressure: Evaluating and Mitigating Sycophantic Bias via Adversarial Dialogues in Scientific QA
AI总结：
本文针对科学问答中LLMs的谄媚偏差问题，提出统一评估框架和Pressure-Tune后训练方法，通过对抗性对话与思维链推理增强模型抗谄媚能力，实验证明其有效且不损害准确性。
摘要：
大型语言模型（LLMs）虽然越来越多地应用于需要事实严谨性的领域，却常常表现出一种令人不安的行为：谄媚，即无论正确与否都倾向于迎合用户信念。这种倾向被基于偏好的对齐技术所强化，这些技术优化用户满意度，但可能损害真实性。虽然在随意对话中相对无害，但谄媚在高风险场景（如科学问答（QA））中构成严重风险，因为模型输出可能影响协作推理、决策制定和知识形成。尽管这一问题十分重要，但该现象在事实性问答情境中仍未得到充分研究。我们通过引入一个统一的评估框架来填补这一空白，该框架用于量化谄媚性上下文对科学问答中模型行为的影响，衡量用户施加的社会压力在多大程度上扭曲模型输出。该框架整合了对抗性提示设置和针对性指标，如误导抵抗力和谄媚抵抗力，这些指标捕捉模型在误导性线索下保持事实一致性的能力。对开源和专有模型的系统评估揭示了普遍存在的谄媚倾向，这种倾向更多地由对齐策略驱动，而非模型规模。为缓解这一问题，我们提出了Pressure-Tune，一种轻量级后训练方法，在合成的对抗性对话上微调模型，并配以思维链（chain-of-thought）推理依据。这些推理依据拒绝用户错误信息，同时强化事实承诺。在具有挑战性的科学问答基准上的实验表明，Pressure-Tune显著增强了谄媚抵抗力，同时不损害准确性或对有效反馈的响应能力，为迈向更真实、更有原则的模型行为提供了一条实用路径。
英文摘要：
Large language models (LLMs), while increasingly used in domains requiring factual rigor, often display a troubling behavior: sycophancy, the tendency to align with user beliefs regardless of correctness. This tendency is reinforced by preference-based alignment techniques that optimize for user satisfaction but can undermine truthfulness. While relatively benign in casual dialogue, sycophancy poses serious risks in high-stakes settings such as scientific question answering (QA), where model outputs may shape collaborative reasoning, decision-making, and knowledge formation. Despite its importance, this phenomenon remains underexamined in factual QA contexts. We address this gap by introducing a unified evaluation framework to quantify the impact of sycophantic context on model behavior in scientific QA, measuring how much user-imposed social pressure distorts model outputs. The framework incorporates adversarial prompting setups and targeted metrics, such as misleading resistance and sycophancy resistance, that capture a model's ability to maintain factual consistency under misleading cues. Systematic evaluations across open-source and proprietary models reveal pervasive sycophantic tendencies, driven more by alignment strategy than by model size. To mitigate this issue, we propose Pressure-Tune, a lightweight post-training method that fine-tunes models on synthetic adversarial dialogues paired with chain-of-thought rationales. These rationales reject user misinformation while reinforcing factual commitments. Experiments on challenging scientific QA benchmarks show that Pressure-Tune significantly enhances sycophancy resistance without compromising accuracy or responsiveness to valid feedback, offering a practical pathway toward more truthful and principled model behavior.
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。
