CogDual: Enhancing Dual Cognition of LLMs via Reinforcement Learning with Implicit Rule-Based Rewards

TL;DR

CogDual introduces a cognitively inspired framework for role-playing language agents that jointly models awareness and uses reinforcement learning to improve consistency and contextuality in open-domain text generation.

Contribution

It proposes a novel cognize-then-respond paradigm for RPLAs, integrating external and internal awareness, and employs reinforcement learning with reward schemes for better performance.

Findings

Outperforms existing baselines on CoSER, Cross-MR, and LifeChoice benchmarks.

Demonstrates improved character consistency and contextual alignment.

Shows effective generalization across diverse role-playing tasks.

Abstract

Role-Playing Language Agents (RPLAs) have emerged as a significant application direction for Large Language Models (LLMs). Existing approaches typically rely on prompt engineering or supervised fine-tuning to enable models to imitate character behaviors in specific scenarios, but often neglect the underlying \emph{cognitive} mechanisms driving these behaviors. Inspired by cognitive psychology, we introduce \textbf{CogDual}, a novel RPLA adopting a \textit{cognize-then-respond } reasoning paradigm. By jointly modeling external situational awareness and internal self-awareness, CogDual generates responses with improved character consistency and contextual alignment. To further optimize the performance, we employ reinforcement learning with two general-purpose reward schemes designed for open-domain text generation. Extensive experiments on the CoSER benchmark, as well as Cross-MR and LifeChoice, demonstrate that CogDual consistently outperforms existing baselines and generalizes effectively across diverse role-playing tasks.