共情是可引导的但具有多轴性：LLM中的机制几何与人格效应
Empathy Is Steerable but Multi-Axial: Mechanism Geometry and Persona Effects in LLMs
浏览论文内容
中文总结 AI 辅助
本研究基于EPITOME框架发现，大语言模型中的共情可通过激活引导部分控制，但方向间不可完全分离，人格提示仅被捕获约3%，表明共情具有多轴性，需超越单方向机制进行控制。
中文摘要 AI 辅助
激活引导已被用于控制诚实、拒绝和谄媚等特质，然而支持性共情是在多个维度上评估的，这些维度未必对应于可独立控制的激活方向。利用EPITOME框架（该框架将支持性共情分解为情绪反应、解释和探索三个成分），我们研究了三个指令微调的大语言模型，并探究从这些标签中得到的候选方向是否产生可区分的干预效果，抑或共享结构，以及人格提示如何与这些方向交互。我们发现，对比激活加法产生一种稳定的中层干预，该干预一致地跨模型改变EPITOME代理分数，将共情分析推进到超越响应级评分。然而，恢复出的方向仅部分可分离：引导一个方向会引发非目标偏移，而手工设计的提示会改变共情轮廓而非隔离单一维度。人格提示显著改变EPITOME分数，但配对激活偏移分解表明，在第15层，恢复出的子空间仅捕获了人格诱导的平方激活偏移幅度的大约3%。基于这种EPITOME定义，表达出的共情是可引导的但具有多轴性，且控制人格条件化的共情需要针对超出单个机制方向的结构。
英文摘要
Activation steering has been used to control traits such as honesty, refusal, and sycophancy, yet supportive empathy is evaluated along multiple dimensions that need not correspond to independently controllable activation directions. Using the EPITOME framework, which decomposes supportive empathy into Emotional Reactions, Interpretations, and Explorations, we study three instruction-tuned LLMs and ask whether candidate directions derived from these labels produce distinguishable intervention effects or instead share structure, and how persona prompts interact with those directions. We find that contrastive activation addition yields a stable middle-layer intervention that consistently shifts the EPITOME proxy scores across models, moving empathy analysis beyond response-level scoring. However, the recovered directions are only partially separable: steering one direction induces off-target shifts, and hand-crafted prompting shifts the empathy profile rather than isolating a single dimension. Persona prompts substantially change EPITOME scores, but a paired activation-shift decomposition shows that the recovered subspace captures only approximately 3 percent of persona-induced squared activation-shift magnitude at layer 15. Under this EPITOME-based definition, expressed empathy is steerable but multi-axial, and controlling persona-conditioned empathy requires targeting structure beyond individual mechanism directions.
发表机构
- Ajou University（亚洲大学）
机构由 AI 辅助整理，请以论文原文为准。
