Memory-Driven Role-Playing: Evaluation and Enhancement of Persona Knowledge Utilization in LLMs

TL;DR

This paper introduces a memory-driven paradigm for LLM role-playing, emphasizing autonomous recall of persona knowledge, and presents evaluation tools and benchmarks to improve and diagnose model performance in maintaining consistent character portrayal.

Contribution

It proposes a novel memory-driven framework, evaluation methods, and a bilingual benchmark for assessing and enhancing persona knowledge utilization in LLMs during role-playing.

Findings

Small models with MRPrompt match larger models' performance.

Memory improvements lead to better response quality.

The paradigm provides comprehensive diagnostic capabilities.

Abstract

A core challenge for faithful LLM role-playing is sustaining consistent characterization throughout long, open-ended dialogues, as models frequently fail to recall and accurately apply their designated persona knowledge without explicit cues. To tackle this, we propose the Memory-Driven Role-Playing paradigm. Inspired by Stanislavski's "emotional memory" acting theory, this paradigm frames persona knowledge as the LLM's internal memory store, requiring retrieval and application based solely on dialogue context, thereby providing a rigorous test of depth and autonomous use of knowledge. Centered on this paradigm, we contribute: (1) MREval, a fine-grained evaluation framework assessing four memory-driven abilities - Anchoring, Recalling, Bounding, and Enacting; (2) MRPrompt, a prompting architecture that guides structured memory retrieval and response generation; and (3) MRBench, a bilingual (Chinese/English) benchmark for fine-grained diagnosis. The novel paradigm provides a comprehensive diagnostic for four-staged role-playing abilities across 12 LLMs. Crucially, experiments show that MRPrompt allows small models (e.g., Qwen3-8B) to match the performance of much larger closed-source LLMs (e.g., Qwen3-Max and GLM-4.7), and confirms that upstream memory gains directly enhance downstream response quality, validating the staged theoretical foundation.