Work

MENTOR: Mitigating Identity Drift in Dynamic Role-Playing via Dual-Chain Structured Memory

Year: 2026

Type: conference-paper

Abstract: Long-context LLM agents increasingly serve multiple users or personas within a single session, requiring stable identity and knowledge boundaries under frequent switching.We identify a common failure mode, identity drift, where models conflate user-specific states and leak information across roles.On BEAM-SWITCH, a benchmark for controlled multiuser switching, performance consistently degrades as switching intensifies, even when responses remain fluent and locally coherent.We propose MENTOR, a cognitive architecture that mitigates identity drift without finetuning.MENTOR uses a Dual-Chain Memory Mechanism: a Global Chain (G) for long-term event logging and isolated Role Chains (R r ) as per-role working memories, supported by a semantic Knowledge Graph (K) that filters and verifies role-admissible information before generation.Across six LLM families, MENTOR improves the overall score (Avg) from 0.46 to 0.75 on average (+0.29 absolute), with substantial gains in identity adherence and knowledge fidelity.

Source: Findings of the Association for Computational Linguistics: ACL (Findings of ACL)

DOI: 10.18653/v1/2026.findings-acl.1046

Authors: Zhu Zhuoning, Xingyu Gao, Hailong Shi

Institutions: Chinese Academy of Sciences, Institute of Microelectronics, University of Chinese Academy of Sciences

Language: English

Topic: Speech and dialogue systems

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: National Natural Science Foundation of China, National Science and Technology Major Project

Award: G1333402700