Search papers, labs, and topics across Lattice.
The paper introduces MUSE, a multi-domain Chinese user simulation framework that enhances persona consistency and realism through iterative profile optimization and rubric-guided reinforcement learning. Iterative Profile Self-Evolution (IPSE) refines user profiles by comparing simulated dialogues with real human interactions. Role-Reversal Supervised Fine-Tuning and rubric-based reward models further improve response realism and long-horizon behavioral consistency, leading to superior performance compared to existing baselines.
Achieve more human-like and consistent user simulation in Chinese across multiple domains by iteratively refining user profiles with real dialogue data and aligning behavior with a rubric-guided reward model.
User simulators are essential for the scalable training and evaluation of interactive AI systems. However, existing approaches often rely on shallow user profiling, struggle to maintain persona consistency over long interactions, and are largely limited to English or single-domain settings. We present MUSE, a multi-domain Chinese user simulation framework designed to generate human-like, controllable, and behaviorally consistent responses. First, we propose Iterative Profile Self-Evolution (IPSE), which gradually optimizes user profiles by comparing and reasoning discrepancies between simulated trajectories and real dialogue behaviors. We then apply Role-Reversal Supervised Fine-Tuning to improve local response realism and human-like expression. To enable fine-grained behavioral alignment, we further train a specialized rubric-based reward model and incorporate it into rubric-guided multi-turn reinforcement learning, which optimizes the simulator at the dialogue level and enhances long-horizon behavioral consistency. Experiments show that MUSE consistently outperforms strong baselines in both utterance-level and session-level evaluations, generating responses that are more realistic, coherent, and persona-consistent over extended interactions.
