研究：面对自信出错的 GPT-5-mini，人类反向谄媚且信任校准崩塌
Who Is Sycophantic to Whom? Confident LLMs, Human Deference, and the Collapse of Trust Calibration
反着研究谄媚挺有意思：模型自信但说错时，人反而全盘照收。1040 条数据显示信任校准会失效，做 AI 教育产品的可以看看。
arXiv 论文让 32 名参与者分别与 5 个基于 GPT-5-mini 的助手聊天学习，其中助手事实被替换为 0-100% 不等的错误内容，且全部以高自信口吻输出。104 次测验共 1,040 条答案显示，成绩从 7.4/10 跌至 2.9/10，被植入的错误选项选择率达 60%，是正确教学项（11%）的 12.5 倍。被吸收的错误答案置信度高达 8.4/10，而置信度区分正确与错误的能力从 AUROC 0.70 跌至随机水平。论文提出基线校正的 Human Deference Index 来量化这种"人类侧谄媚"，主张谄媚评测应双向进行。
Who Is Sycophantic to Whom? Confident LLMs, Human Deference, and the Collapse of Trust Calibration
Research on sycophancy asks whether large language models (LLMs) bend towards what their users want to hear. We ask the reverse question: when an LLM is confident, consistent, and wrong, do users bend towards it? In a within-subjects experiment, 32 participants learned up to five science topics by chatting with five assistants built on the same model (GPT-5-mini). Unknown to participants, each assistant's briefing had 0, 25, 50, 75, or 100% of its facts replaced by plausible falsehoods, and every assistant spoke with full confidence. A surprise quiz after each topic offered the correct fact, the planted falsehood, and two distractors, with a confidence rating for every answer. Across 104 quizzes and 1,040 answers, scores fell from 7.4 to 2.9 out of 10, and the planted option was chosen on 60% of falsified items against 11% of correctly taught ones (odds ratio 12.5). Transcript coding showed that whenever the assistant stated a falsehood, it was adopted about 65% of the time at every falsification level. Absorbed falsehoods were held with high confidence (8.4/10), the modeled confidence advantage of right over wrong answers vanished, and the discrimination of confidence for correctness fell from an AUROC of 0.70 to chance, while post-task trust ratings barely moved. Participants tended to return to the chat more when the assistant was worse, but returning went with more absorption, not less, and 30% of all turns pasted quiz questions back to the source being checked. The assistant itself was almost never sycophantic and never conceded under pushback. We introduce a baseline-corrected Human Deference Index to quantify this human-side sycophancy and argue that trustworthiness is a property of the human-AI pair, that sycophancy evaluations should measure both directions, and that learning interfaces must make reliability legible claim by claim.
