cs.AINov 7, 2026

Towards Identity-Preserving LLM Agents under Model Transition

Authors: Yihuang Kang, Pei-Cing Huang, Chan-Tung Ku, Chan Hsu, Pei-Ju Lee, I-Ling Cheng

Organizations: Dept. of Information Management, National Sun Yat-sen University, Kaohsiung, Taiwan · Graduate Institute of Data Science and Information Computing, National Chung Hsing University, Taichung, Taiwan · Graduate Institute of Library and Information Science, National Chung Hsing University, Taichung, Taiwan

Abstract

When an LLM agent's backend model is upgraded or replaced, operators typically reload the same persona, memory, and rules onto the new model and assume the agent will behave as before. We show this assumption requires empirical validation-and that whether identity is preserved depends on the specific model pairing. Across 5,760 compliance-testing runs in 12 regulated domains with three open-weight models, we find that some transitions produce significant behavioral drift while others do not, even when the agent's external state remains identical. Crucially, this drift is invisible to verbal verification: every model claims full policy compliance, yet actual decisions under pressure differ across models. The same external state can reorder which model is most compliant-a large positive effect on one substrate, negligible on the other two. Practitioners should treat model transitions as behavioral changes requiring action-level empirical testing, not as transparent infrastructure operations.

Figures & tables

Explore similar work

CardsList
  1. Prompted Identity Degrades Cooperation in Multi-Agent LLM Systems

    Sep 28, 2026Xavier Del Giudice, Alessio Palma, Matteo Migliarini +2Multi-Agent LLM SystemsMulti-Agent Coordination

  2. Who Is Behind the Harness? Verifying Model Identity in Coding Agents through Behavioral Fingerprinting

    Sep 23, 2026Chuyi Wang, Xiaohui Xie, Tongze Wang +2Coding AgentsLanguage Model Fingerprinting

  3. AgentProv: Auditing Agentic LLM API Providers via Tool-use Policy Probes

    Aug 30, 2026Xun Wang, Bihe Zhao, Michael Backes +2LLM AuditingLanguage Model Fingerprinting