Enhancing Persona Following at Decoding Time via Dynamic Importance Estimation for Role-Playing Agents

2026-03-01

AI Summary

This work addresses the challenge that existing role-playing language agents struggle to maintain behavioral consistency with their character profiles in dynamic scenarios, primarily due to insufficient modeling of the contextual dependence of character attributes. To this end, the authors propose a novel decoding-stage approach that dynamically estimates the importance of character attributes and guides text generation accordingly. The method introduces, for the first time, an unsupervised character importance estimation mechanism grounded in the Cognitive-Affective Personality System, integrated with a Personality-guided Inference Alignment (PIA) paradigm and a weighted multi-objective reward decoding strategy. Experimental results demonstrate that the proposed approach significantly outperforms baseline models in both utterance consistency and behavioral fidelity, thereby validating the effectiveness and novelty of dynamic character management during generation.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Natural Language Processing: Generation
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search

Abstract

The utility of Role-Playing Language Agents in sociological research is growing alongside the adoption of Large Language Models. For realism in social simulation, these agents must adhere to their personas defined by character profiles, yet existing strategies-static prompt engineering or costly fine-tuning-fail to adapt personas to dynamic scenarios. Psychological theories, such as the Cognitive-Affective Personality Systems, provide a crucial explanation for this failure: a persona's influence on behavior is not static but varies with the scenarios. This context-dependence highlights the critical need for adaptive persona management. To address this gap, we propose a novel, theory-driven method that dynamically estimates context-dependent persona importance and integrates it into weighted reward-guided decoding, enabling inference-time persona following. Specifically, we introduce the Persona Dynamic Decoding (PDD) framework, which consists of two key components: (1) Persona Importance Estimation (PIE) module, which dynamically quantifies the contextual importance of persona attributes without requiring ground-truth supervision; and (2) Persona-Guided Inference-Time Alignment (PIA) paradigm, which leverages these importance scores to construct weighted multi-objective rewards and modulate generation probabilities during inference. Extensive experiments show the effectiveness of our method in utterance consistency and behavioral fidelity.

Problem

• persona following
• role-playing agents
• context-dependent behavior
• adaptive persona management
• social simulation

Innovation

• Dynamic Persona Importance
• Inference-Time Alignment
• Role-Playing Agents
• Weighted Reward-Guided Decoding
• Context-Dependent Persona