大语言模型中认知结构的机制解码
Mechanistic Decoding of Cognitive Constructs in Large Language Models
AI总结：
本文提出基于Representation Engineering的Cognitive Reverse-Engineering框架，分析社会比较嫉妒的认知机制，揭示模型内部对嫉妒的结构化编码，并展示如何机械检测和抑制有毒情绪状态。
摘要：
尽管大型语言模型（LLMs）展现出日益复杂的情感能力，但它们处理复杂情绪的内部机制仍不明确。现有可解释性方法往往将模型视为黑箱或聚焦于基础情绪的粗粒度分析，导致更复杂情绪状态的认知结构研究不足。为此，我们提出基于表示工程（RepE）的认知逆向工程框架，分析社会比较嫉妒。通过结合评价理论与子空间正交化、回归加权以及双向因果引导，我们隔离并量化了嫉妒的两个心理前因，即比较对象的优越性和领域自我定义相关性，并检验它们对模型判断的因果影响。在八个LLM（Llama、Qwen和Gemma家族）上的实验表明，模型本征编码嫉妒为这些构成因素的结构化线性组合。其内部表示与人类心理构念广泛一致，将优越性视为基础触发因素，将相关性视为最终强度乘数。我们的框架还表明，有毒情绪状态可以机械检测并手术抑制，暗示了在多智能体环境中实现表示监控和干预的可能路径。
英文摘要：
While Large Language Models (LLMs) demonstrate increasingly sophisticated affective capabilities, the internal mechanisms by which they process complex emotions remain unclear. Existing interpretability approaches often treat models as black boxes or focus on coarse-grained basic emotions, leaving the cognitive structure of more complex affective states underexplored. To bridge this gap, we propose a Cognitive Reverse-Engineering framework based on Representation Engineering (RepE) to analyze social-comparison jealousy. By combining appraisal theory with subspace orthogonalization, regression-based weighting, and bidirectional causal steering, we isolate and quantify two psychological antecedents of jealousy, Superiority of Comparison Person and Domain Self-Definitional Relevance, and examine their causal effects on model judgments. Experiments on eight LLMs from the Llama, Qwen, and Gemma families suggest that models natively encode jealousy as a structured linear combination of these constituent factors. Their internal representations are broadly consistent with the human psychological construct, treating Superiority as the foundational trigger and Relevance as the ultimate intensity multiplier. Our framework also demonstrates that toxic emotional states can be mechanically detected and surgically suppressed, suggesting a possible route toward representational monitoring and intervention for AI safety in multi-agent environments.
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。
