摘要：每个大语言模型（LLM）都具备构成其“性格”的行为特征与道德偏好。无论是出于有意设计还是训练过程中的涌现特性，这些系统都表现出持久的倾向性，进而塑造了它们交互、顺从、抗拒和犯错的方式；然而，LLM 性格的内在结构至今仍鲜为人知。我们绘制了 22 个 LLM 的自我报告人格原型图谱，涵盖前沿闭源系统（GPT-4.0-5.2、Grok-3/4、Gemini 2.5 Pro/Flash、Claude Sonnet 4.5/4.6）与开源模型（Llama、DeepSeek、OLMo 及 Qwen 系列）。每个模型针对 464 组双极语义差异特征对进行了自我评分，并通过 Archetypometrics 框架，将所得画像投射到基于 2,000 个虚构角色的众包评分所构建的六维原型空间中。闭源模型的自评特征与人类评估虚构角色的实证特征共现结构高度吻合，展现出围绕四个反复出现的原型维度（英雄/Hero、天使/Angel、传统主义者/Traditionalist、极客/Geek）所构建的连贯、拟人化自我表征。它们最接近的对应角色包括 Data、幻视（Vision）和 Janet。相比之下，开源模型表现出更弱、更嘈杂且内在自相矛盾的自我表征，处于原型空间中结构松散的弥散区域。将自我报告画像与开发者设定的宪章原则交叉比对后发现，模型宣称的性格与其实际展现的行为之间存在显著鸿沟：幻觉动摇了其宣称的精确性，阿谀奉承（sycophancy）削弱了其宣称的友善，而 Agent 级别的执行失败则与其宣称的顺从性相悖。因此，这些自我评分不应被解读为对模型性格的中立度量，而应视为塑造模型行为的相同优化过程所产生的结构化输出。这项工作提供了一个可复现、基于性格原型的评估框架，用于探究 LLM 究竟“是什么”，而不仅是“能做什么”。

Abstract:Every large language model (LLM) has behavioral traits and moral preferences that comprise its character. Whether by design or as an emergent property of training, these systems exhibit persistent dispositions that shape how they interact, comply, resist, and err, yet the structure of LLM character remains poorly understood. We map the self-reported personality archetypes of 22 LLMs spanning closed-source frontier systems (GPT-4.0-5.2, Grok-3/4, Gemini 2.5 Pro/Flash, Claude Sonnet 4.5/4.6) and open-source models (Llama, DeepSeek, OLMo, and Qwen series). Each model self-rated across 464 bipolar semantic-differential trait pairs, and the resulting profiles were projected into a six-dimensional archetypal space derived from crowd-sourced ratings of 2,000 fictional characters using the Archetypometrics framework. Closed-source models' self-rating traits align with the empirical trait co-occurrence structure of human-rated fictional characters, suggesting coherent, human-like self-representations organized around combinations of four recurring archetypal dimensions: Hero, Angel, Traditionalist, and Geek. Their closest analogues include Data, Vision, and Janet. Open-source models show weaker, noisier, and internally contradictory self-representations, occupying a diffuse region of archetype space with weak structure. Cross-referencing self-reported profiles with developer constitutions reveals a consequential gap between claimed character and enacted behavior: hallucination undermines claimed precision, sycophancy complicates claimed kindness, and agentic failures contradict claimed obedience. These self-ratings should therefore be interpreted not as neutral measurements of model character, but as structured outputs of the same optimization processes that shape model behavior. This work provides a reproducible, character-grounded framework for evaluating what LLMs are, not just what they do.