Most LLM Conformity Needs No Speaker: Measuring the Speaker-Free Floor in Peer-Pressure Benchmarks

大多数大模型从众行为无需“发言者”：衡量同侪压力基准测试中的“无发言者基准线”

Abstract: LLM conformity is often used to describe cases where a model changes a correct answer toward a peer or group response. We show that most of this apparent conformity survives even after the peer is removed. The reason is a confound: standard conformity prompts mix two cues at once, the presence of a speaker and the repeated wrong answer itself. Existing benchmarks vary these cues together, so they cannot tell how much of the revision actually depends on the speaker.

摘要： 大模型（LLM）的从众行为通常被用来描述模型将正确答案更改为与同侪或群体回答一致的情况。我们研究发现，即使移除了“同侪”这一因素，大部分这种表面的从众行为依然存在。其原因是存在混淆变量：标准的从众测试提示词同时混合了两个线索，即发言者的存在以及重复出现的错误答案本身。现有的基准测试将这些线索同时变动，因此无法判断模型修改答案的行为在多大程度上真正依赖于发言者。

We introduce a no-source condition: the same asserted answer with the explicit speaker removed. Across six open-weight LLMs and seven QA and reasoning datasets, this condition alone causes harmful revision in 66.5% of initially correct cases, compared with 10.3% under a plain re-ask. The effect also remains when the repeated answer is paraphrased and when answer options are hidden in an open-ended setting.

我们引入了一种“无来源”条件：在移除明确发言者的情况下，提供相同的断言答案。在六个开源权重的大模型和七个问答及推理数据集上，仅此条件就导致了 66.5% 的初始正确案例被错误修改，而简单的“再次询问”仅导致 10.3% 的错误修改。当重复的答案被改写，或者在开放式问答中隐藏选项时，这种效应依然存在。

Source framing mainly modulates this floor: expert-panel framing raises it, while minimal person labels do not reliably raise it. When models flip, they are usually confidently wrong, and simple recalibration does not recover the original answer. Source attribution still matters, but it should be measured as an increment above this speaker-free floor. The methodological lesson is that conformity benchmarks should first measure what remains after the speaker is removed; without this step, benchmarks may mistake repeated text for social influence.

来源的框架设定主要调节了这一基准线：专家小组的框架设定会提高从众率，而简单的个人标签则不会可靠地提高它。当模型改变答案时，它们通常表现得“自信地错误”，且简单的重新校准无法恢复原始答案。来源归因仍然很重要，但应将其衡量为高于这一“无发言者基准线”的增量。方法论上的启示是：从众基准测试应首先测量移除发言者后剩余的从众行为；若不进行这一步骤，基准测试可能会将重复出现的文本误认为是社会影响力。