AI 中文总结
该研究探究LLMs对语境真值的编码，发现其会维持语境真值的线性表征，伙伴断言可改变命题真值表征，还区分出两种谄媚形式并量化了其出现频率差异。
AI 中文摘要
已有研究表明，大型语言模型（LLMs）会在激活空间中沿线性方向编码事实命题的真值。目前尚不清楚这些表征如何延伸至语境真值：即真值由上下文证据而非世界知识决定的命题。本文表明，LLMs会维持语境真值的线性表征，该表征在结构不同的输出策略间仍持续存在，即便输出不要求模型判定命题真值，也会通过引导实验提供因果证据。利用需两个LLMs维持共同语境的协作视觉-语言任务的对话记录，本文发现，即便LLM已有足够证据判定命题真值，其关于某命题的真值表征仍会被伙伴对该命题的断言显著影响。研究还发现，处于决策边界附近的命题更易因伙伴断言改变真值。通过将表征与输出分离，可区分仅靠输出行为无法识别的两种谄媚形式：模型可能接纳错误命题但仍将其表征为错误，或跨越边界改变表征；当模型通过明确重述错误主张表示同意时，后者出现频率是隐含同意时的2.59倍。
英文摘要
Prior work has shown that LLMs encode the truth of factual propositions along linear directions in activation space. It's unclear how these representations extend to contextual truth: propositions whose truth is determined by in-context evidence rather than world knowledge. We show that LLMs maintain a linear representation of contextual truth that persists across structurally different output policies, even when the output doesn't require the model to determine a proposition's truth, and show causal evidence via steering experiments. Using the transcripts from a collaborative vision-language task that requires two LLMs to maintain a shared common ground, we show that truth representations of a proposition are significantly swayed by partner assertions about that proposition, even when the LLM has enough evidence to determine its truth. We find evidence that propositions near the decision boundary are more susceptible to having their truth shifted through partner assertions. Separating representation from output distinguish two forms of sycophancy that output behavior alone cannot: the model may accommodate a false proposition while continuing to represent it as false, or shift its representation across the boundary. The latter is 2.59x more common when the model agrees by restating the false claim explicitly than when it agrees implicitly.
