mlsanityData to 7 Oct 2026

The Chameleon's Limit: Investigating Persona Collapse and Homogenization in Large Language Models

arXiv 2604.24698 ↗First posted 27 Apr 2026cs.CLcited by 6

Not on the map. Why?

Abstract (arXiv)

Applications based on large language models (LLMs), such as multi-agent simulations, require population diversity among agents. We identify a pervasive failure mode we term Persona Collapse: agents each assigned a distinct profile nonetheless converge into a narrow behavioral mode, producing a homogeneous simulated population. To quantify persona collapse, we propose a framework that measures how much of the persona space a population occupies (Coverage), how evenly agents spread across it (Uniformity), and how rich the resulting behavioral patterns are (Complexity). Evaluating ten LLMs on personality simulation (BFI-44), moral reasoning, and self-introduction, we observe persona collapse along two axes: (1) Dimensions: a model can appear diverse on one axis yet structurally degenerate on another, and (2) Domains: the same model may collapse the most in personality yet be the most diverse in moral reasoning. Furthermore, item-level diagnostics reveal that behavioral variation tracks coarse demographic stereotypes rather than the fine-grained individual differences specified in each persona. Counter-intuitively, the models achieving the highest per-persona fidelity consistently produce the most stereotyped populations. We release our toolkit and data to support population-level evaluation of LLMs.

Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Builds on 33 of its 49 references that are in the map. First posted on arXiv 27 Apr 2026.

Builds on

33 of 49 references.

References · 33

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  3. 2025Qwen3 Technical Reportcited by 8,195
  4. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,462
  5. 2023Generative Agents: Interactive Simulacra of Human Behaviorcited by 3,500
  6. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
  7. 2020Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hyperspherecited by 1,393
  8. 2023Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debatecited by 995
  9. 2019Improved Precision and Recall Metric for Assessing Generative Modelscited by 925
  10. 2020Reliable Fidelity and Diversity Metrics for Generative Modelscited by 425
  11. 2023Character-LLM: A Trainable Agent for Role-Playingcited by 335
  12. 2023Role-Play with Large Language Modelscited by 290
21 more
  1. 2023PersonaLLM: Investigating the Ability of Large Language Models to Express Personality Traitscited by 229
  2. 2023Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMscited by 191
  3. 2023InCharacter: Evaluating Personality Fidelity in Role-Playing Agents through Psychological Interviewscited by 177
  4. 2024Investigating Cultural Alignment of Large Language Modelscited by 140
  5. 2023Intrinsic Dimension Estimation for Robust Detection of AI-Generated Textscited by 111
  6. 2024On the Resilience of LLM-Based Multi-Agent Collaboration with Faulty Agentscited by 92
  7. 2024Roleplay-doh: Enabling Domain-Experts to Create LLM-simulated Patients via Eliciting and Adhering to Principlescited by 72
  8. 2023When "A Helpful Assistant" Is Not Really Helpful: Personas in System Prompts Do Not Improve Performances of Large Language Modelscited by 50
  9. 2025The Prompt Makes the Person(a): A Systematic Evaluation of Sociodemographic Persona Prompting for Large Language Modelscited by 40
  10. 2025Consistently Simulating Human Personas with Multi-Turn Reinforcement Learningcited by 35
  11. 2024LLMs generate structurally realistic social networks but overestimate political homophilycited by 31
  12. 2025Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Modelscited by 30
  13. 2025CoSER: A Comprehensive Literary Dataset and Framework for Training and Evaluating LLM Role-Playing and Persona Simulationcited by 26
  14. 2025LLM Probability Concentration: How Alignment Shrinks the Generative Horizoncited by 26
  15. 2025Synthetic Socratic Debates: Examining Persona Effects on Moral Decision and Persuasion Dynamicscited by 19
  16. 2025VisBias: Measuring Explicit and Implicit Social Biases in Vision Language Modelscited by 14
  17. 2025The PIMMUR Principles: Ensuring Validity in Collective Behavior of LLM Societiescited by 12
  18. 2025Humanizing Machines: Rethinking LLM Anthropomorphism Through a Multi-Level Framework of Designcited by 11
  19. 2026MASCOT: Multi-Agent Socio-Collaborative Companion Systemscited by 7
  20. 2026Knowing But Not Doing: Convergent Morality and Divergent Action in LLMscited by 6
  21. 2026Sentipolis: Emotion-Aware Agents for Social Simulationscited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 6.

6 papers.

  1. 2026One Frozen Simulator Is Not Enough: Simulator Collapse in Multi-Agent RLcited by 1
  2. 2026BOUNDARY_SYNC: Measuring Communication-Induced Representational Coupling in Multi-Agent LLM Systemscited by 0
  3. 2026Co-design of LLM-based preference agents: participation may drive overtrustcited by 0
  4. 2026When Cosine Similarity Fails to Reflect Linearly Accessible Structure in Dialogue Modelscited by 0
  5. 2026How Far Do Persona Effects Generalize in Language Models?cited by 0
  6. 2026Back in Style: A Sociolinguistic Approach to Authoring and Measuring Persona Fidelity in User Simulationcited by 0