mlsanityData to 14 Oct 2026

MicroVerse: An Instrument for Measuring Self-Authored Identity Drift in Long-Horizon Multi-Agent Language-Model Simulations

arXiv 2608.15844 ↗First posted 16 Aug 2026cs.CLcited by 1

On the map: LLM agents and tool use → Social simulation and multi-agent debate

Abstract (arXiv)

Long-horizon, multi-agent language model (LM) simulations are widely proposed for studying social behavior, yet instruments to measure whether persona-conditioned agents maintain identity fidelity under sustained pressure are lacking. We present MicroVerse, a behavioral-science instrument that measures identity drift in generative agents. Agents carry an immutable "soul file" (core values, moral boundaries, personality, goals) and inhabit a resource-scarce 50 x 50 environment where water is a non-respawning survival constraint. Scarcity is operationalized via a per-tick existence-cost gradient. The eight-verb action space maps directly to moral boundaries (trade, talk, attack, scavenge). Using a three-layer memory architecture, agents periodically revise a mutable current identity against their immutable original soul via importance-triggered reflection. To mitigate survivor bias, MicroVerse decouples measurement from behavior using uniform longitudinal engine snapshots every N ticks alongside a forced-end snapshot of all living and dead agents. Identity drift is scored offline using a paraphrase-aware, value-anchored, multi-register diff rather than raw cosine similarity. We evaluate the instrument via a controlled seed run (n = 25) and a reflection-threshold sweep (thresholds {40, 80, 150}) to determine if drift dynamics are gate artifacts or threshold-robust properties. We report two primary findings: (1) Anti-self-deception emerges unprompted as the single largest semantic category of identity modification (27 of 111 added boundaries, 24%). (2) The system is threshold-robust; lower gates accelerate and increase revision frequency but preserve drift direction. All empirical results are strictly preliminary existence proofs and effect shapes (one model, one seed per arm, n = 25) rather than statistical significance claims.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Social simulation and multi-agent debate, one of 16 directions in LLM agents and tool use. Builds on 39 of its 45 references that are in the map. First posted on arXiv 16 Aug 2026.

Builds on

39 of 45 references.

Closest ancestors · 7

Same direction, newest first

  1. 2026SPASM: Stable Persona-driven Agent Simulation for Multi-turn Dialogue Generation10 Apr
  2. 2025Do Role-Playing Agents Practice What They Preach? Belief-Behavior Consistency in LLM-Based Simulations of Human Trust2 Jul
  3. 2025AgentSociety: Large-Scale Simulation of LLM-Driven Generative Agents Advances Understanding of Human Behaviors and Society12 Feb
  4. 2024LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals15 Nov
  5. 2024SocialBench: Sociality Evaluation of Role-Playing Conversational Agents20 Mar
  6. 2023Generative agent-based modeling with actions grounded in physical, social, or digital space using Concordia6 Dec
  7. 2023Generative Agents: Interactive Simulacra of Human Behavior7 Apr

Other references · 32

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
  3. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  4. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,276
  5. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 3,050
  6. 2023AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversationcited by 1,988
  7. 2023MemGPT: Towards LLMs as Operating Systemscited by 1,389
  8. 2023Towards Understanding Sycophancy in Language Modelscited by 1,040
  9. 2018Personalizing Dialogue Agents: I have a dog, do you have pets too?cited by 828
  10. 2023Cognitive Architectures for Language Agentscited by 429
  11. 2023SOTOPIA: Interactive Evaluation for Social Intelligence in Language Agentscited by 367
  12. 2023Role-Play with Large Language Modelscited by 294
20 more
  1. 2023InCharacter: Evaluating Personality Fidelity in Role-Playing Agents through Psychological Interviewscited by 179
  2. 2024OASIS: Open Agent Social Interaction Simulations with One Million Agentscited by 132
  3. 2023ChatHaruhi: Reviving Anime Character in Reality via Large Language Modelcited by 59
  4. 2024Elaborative Subtopic Query Reformulation for Broad and Indirect Queries in Travel Destination Recommendationcited by 11
  5. 2025MA-DPR: Manifold-aware Distance Metrics for Dense Passage Retrievalcited by 9
  6. 2025Multimodal Item Scoring for Natural Language Recommendation via Gaussian Process Regression with LLM Relevance Judgmentscited by 9
  7. 2025A Simple but Effective Elaborative Query Reformulation Approach for Natural Language Recommendationcited by 8
  8. 2025ChessQA: Evaluating Large Language Models for Chess Understandingcited by 8
  9. 2026Semantic XPath: Structured Agentic Memory Access for Conversational AIcited by 7
  10. 2026Bayesian Active Learning with Gaussian Processes Guided by LLM Relevance Scoring for Dense Passage Retrievalcited by 6
  11. 2024Characteristic AI Agents via Large Language Modelscited by 5
  12. 2026Grounded Chess Reasoning in Language Models via Master Distillationcited by 5
  13. 2026ThinkTwice: Jointly Optimizing Large Language Models for Reasoning and Self-Refinementcited by 5
  14. 2026Evaluating Scene-based In-Situ Item Labeling for Immersive Conversational Recommendationcited by 5
  15. 2023Monte-Carlo Tree Search for Behavior Planning in Autonomous Drivingcited by 4
  16. 2026Goal-Oriented Reasoning for RAG-based Memory in Conversational Agentic LLM Systemscited by 4
  17. 2026SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agentscited by 4
  18. 2026PersonaArena: Dynamic Simulation for Evaluating and Enhancing Persona-Level Role-Playing in Large Language Modelscited by 3
  19. 2026Temporal Order Matters for Agentic Memory: Segment Trees for Long-Horizon Agentscited by 3
  20. 2026Best Friends, Not Forever: Evaluating Long-Horizon Persona Collapse and Behavioral Drift in AI Companionscited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Benchmark Radar: A Living Database and Search Engine for AI Benchmarks and Evaluationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Benchmarking Psychological Dynamics in Generative Agents
  2. 2026Mitigating Identity Essentialism in LLM Agents with Longitudinal Life Trajectories
  3. 2025Agent Identity Evals: Measuring Agentic Identity
  4. 2024Examining Identity Drift in Conversations of LLM Agents
  5. 2026Beyond Preset Identities: How Agents Form Stances and Boundaries in Generative Societies
  6. 2026Time, Identity and Consciousness in Language Model Agents
  7. 2026Large Language Model Agents Are Not Always Faithful Self-Evolvers
  8. 2026Agent Sociodrama: Social-Dynamics-Guided Persona Recasting for LLM Agent Teams