MOA: Multi-Objective Alignment for Role-Playing Agents

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liao, Chonghua, Wang, Ke, Wu, Yuchuan, Li, Ruoran, Huang, Fei, Li, Yongbin
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913052761260032
author Liao, Chonghua
Wang, Ke
Wu, Yuchuan
Li, Ruoran
Huang, Fei
Li, Yongbin
author_facet Liao, Chonghua
Wang, Ke
Wu, Yuchuan
Li, Ruoran
Huang, Fei
Li, Yongbin
contents Role-playing agents (RPAs) require balancing multiple objectives, such as instruction following, persona consistency, and stylistic fidelity, which are not always perfectly aligned across different dimensions. While prior work has primarily relied on supervised fine-tuning or reinforcement learning with scalarized rewards, these approaches do not explicitly address the coordination of multiple reward dimensions during optimization. We present \textbf{MOA} (\textbf{M}ulti-\textbf{O}bjective \textbf{A}lignment), a reinforcement-learning framework that enables multi-dimensional, fine-grained rubric optimization for general RPAs. MOA introduces a novel multi-objective optimization strategy that trains simultaneously on multiple fine-grained rubrics to boost optimization performance. Additionally, to improve both output diversity and generation quality, we employ thought-augmented rollouts with off-policy guidance. Experiments on PersonaGym and RoleMRC show that MOA consistently improves multi-dimensional role-playing performance over supervised and standard RL baselines. Under identical evaluation protocols, an 8B model trained with MOA reaches performance competitive with strong closed-source models across multiple evaluation dimensions. These results suggest that MOA provides a practical framework for training more capable general-purpose role-playing agents.
format Preprint
id arxiv_https___arxiv_org_abs_2512_09756
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MOA: Multi-Objective Alignment for Role-Playing Agents
Liao, Chonghua
Wang, Ke
Wu, Yuchuan
Li, Ruoran
Huang, Fei
Li, Yongbin
Computation and Language
Role-playing agents (RPAs) require balancing multiple objectives, such as instruction following, persona consistency, and stylistic fidelity, which are not always perfectly aligned across different dimensions. While prior work has primarily relied on supervised fine-tuning or reinforcement learning with scalarized rewards, these approaches do not explicitly address the coordination of multiple reward dimensions during optimization. We present \textbf{MOA} (\textbf{M}ulti-\textbf{O}bjective \textbf{A}lignment), a reinforcement-learning framework that enables multi-dimensional, fine-grained rubric optimization for general RPAs. MOA introduces a novel multi-objective optimization strategy that trains simultaneously on multiple fine-grained rubrics to boost optimization performance. Additionally, to improve both output diversity and generation quality, we employ thought-augmented rollouts with off-policy guidance. Experiments on PersonaGym and RoleMRC show that MOA consistently improves multi-dimensional role-playing performance over supervised and standard RL baselines. Under identical evaluation protocols, an 8B model trained with MOA reaches performance competitive with strong closed-source models across multiple evaluation dimensions. These results suggest that MOA provides a practical framework for training more capable general-purpose role-playing agents.
title MOA: Multi-Objective Alignment for Role-Playing Agents
topic Computation and Language
url https://arxiv.org/abs/2512.09756