Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
For researchers and practitioners in role-playing AI, this work addresses the key limitation of supervised fine-tuning by introducing psychologically grounded reasoning and reinforcement learning alignment, leading to more faithful and generalizable character portrayal.
The paper tackles the problem of building general-purpose role-playing agents that faithfully portray characters from natural-language profiles. It proposes Psy-CoT, a psychology-grounded chain-of-thought reasoning framework, and RAPO, a role-aware policy optimization method, achieving state-of-the-art results on CoSER, CharacterBench, and CharacterEval benchmarks, outperforming existing role-playing CoT methods and GRPO across multiple model scales.
Building general-purpose role-playing agents that faithfully portray any character from a natural-language profile remains challenging. The dominant paradigm -- supervised fine-tuning -- encourages behavioral mimicry without deep, human-like internal thought processes, resulting in poor out-of-distribution generalization. Therefore, we propose \textbf{Psy-CoT}, a psychology-grounded chain-of-thought framework that decomposes pre-response reasoning into three role-specific steps -- \emph{Interaction Perception}, \emph{Psychological Empathy}, and \emph{Logical Construction} -- so that the model \emph{thinks dynamically} from the profile rather than merely mimicking surface patterns. While structured reasoning provides a foundation, it alone is insufficient; reinforcement learning is essential to further align the model with character fidelity. However, we observe that under LLM-based reward models, both generic phrases that hack the reward model and genuinely role-specific phrases receive identical gradient signals -- this hacking accumulates over training, misleading the model into treating both as equally optimal choices. To address this, we propose \textbf{Role-Aware Policy Optimization (RAPO)}, which uses profile--token mutual information to weight gradients asymmetrically -- amplifying role-specific tokens under positive advantage while attenuating them under negative advantage. Experiments on CoSER, CharacterBench, and CharacterEval demonstrate that Psy-CoT outperforms existing role-playing CoT methods, and RAPO consistently surpasses GRPO across multiple model scales.