arXiv cs.CL· Yunze Xiao, Vivienne J. Zhang, Chenghao Yang, Weihao Xuan, Jen-tse Huang·· 12 小时前AI 评分42
大语言模型的“变色龙极限”：探究人格坍缩与同质化
The Chameleon's Limit: Investigating Persona Collapse and Homogenization in Large Language Models
AI 导读
研究将 LLM 多智能体模拟中人格响应聚集、无法覆盖人群多样性的现象形式化为“人格坍缩”，并用 Coverage、Uniformity、Complexity 三项指标量化其群体结构。
正文 · 原文
Abstract:Large language models (LLMs) have powered multi-agent simulations such as survey respondents or application users. However, LLMs lack the population-level diversity required by these simulations---responses from LLMs with detailed human profiles cluster into classes, unlike human responses that uniformly fill the entire space. In this paper, we formalize this as Persona Collapse and quantify its population-level structure with three diagnostic metrics: Coverage (how much of the persona space is occupied), Uniformity (how evenly agents spread across it), and Complexity (how rich the resulting behaviors are). Across eight LLMs on personality (BFI-44), moral reasoning, and self-introduction, persona collapse is universal but task-dependent: CoSER-Llama-8B collapses the most in personality while being the most diverse in moral reasoning. Moreover, sampling more personas cannot mitigate the collapse. We further show a fidelity--diversity tradeoff: the models with the highest per-persona fidelity produce the most polarized populations. Finally, persona-targeted fine-tuning further narrows the population instead of diversifying it.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2604.24698 [cs.CL] |
| (or arXiv:2604.24698v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2604.24698 arXiv-issued DOI via DataCite |
Submission history
From: Yunze Xiao [view email]
[v1]
Mon, 27 Apr 2026 17:01:48 UTC (1,773 KB)
[v2]
Fri, 16 Oct 2026 15:30:51 UTC (1,242 KB)
来源：arXiv cs.CL · arxiv.org