CPO: Addressing Reward Ambiguity in Role-playing Dialogue via Comparative Policy Optimization

• URL: http://arxiv.org/abs/2508.09074v1
• Date: Tue, 12 Aug 2025 16:49:18 GMT
• Title: CPO: Addressing Reward Ambiguity in Role-playing Dialogue via Comparative Policy Optimization
• Authors: Xinge Ye, Rui Wang, Yuchuan Wu, Victor Ma, Feiteng Fang, Fei Huang, Yongbin Li,
• Abstract summary: Reinforcement Learning Fine-Tuning (RLFT) has achieved notable success in tasks with objectively verifiable answers.<n>But it struggles with open-ended subjective tasks like role-playing dialogue.<n>Traditional reward modeling approaches, which rely on independent sample-wise scoring, face dual challenges: subjective evaluation criteria and unstable reward signals.<n>Motivated by the insight that human evaluation inherently combines explicit criteria with implicit comparative judgments, we propose Comparative Policy Optimization.
• Score: 53.79487826635141
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: Reinforcement Learning Fine-Tuning (RLFT) has achieved notable success in tasks with objectively verifiable answers (e.g., code generation, mathematical reasoning), yet struggles with open-ended subjective tasks like role-playing dialogue. Traditional reward modeling approaches, which rely on independent sample-wise scoring, face dual challenges: subjective evaluation criteria and unstable reward signals.Motivated by the insight that human evaluation inherently combines explicit criteria with implicit comparative judgments, we propose Comparative Policy Optimization (CPO). CPO redefines the reward evaluation paradigm by shifting from sample-wise scoring to comparative group-wise scoring.Building on the same principle, we introduce the CharacterArena evaluation framework, which comprises two stages:(1) Contextualized Multi-turn Role-playing Simulation, and (2) Trajectory-level Comparative Evaluation. By operationalizing subjective scoring via objective trajectory comparisons, CharacterArena minimizes contextual bias and enables more robust and fair performance evaluation. Empirical results on CharacterEval, CharacterBench, and CharacterArena confirm that CPO effectively mitigates reward ambiguity and leads to substantial improvements in dialogue quality.

Related papers

• Provable and Practical In-Context Policy Optimization for Self-Improvement [49.670847804409874]
  We study test-time scaling, where a model improves its answer through multi-round self-reflection at inference.<n>We introduce In-Context Policy Optimization (ICPO), in which an agent optimize its response in context using self-assessed or externally observed rewards without modifying its parameters.<n>We propose Minimum-Entropy ICPO (ME-ICPO), a practical algorithm that iteratively uses its response and self-assessed reward to refine its response in-context at inference time.
  arXiv  Detail & Related papers  (2026-03-02T00:21:50Z)
• RoleRMBench & RoleRM: Towards Reward Modeling for Profile-Based Role Play in Dialogue Systems [85.16327248973387]
  We develop RoleRM, a reward model trained with Continuous Implicit Preferences (CIP)<n>We show RoleRM surpasses strong open- and closed-source reward models by over 24% on average.<n>Our findings highlight the importance of continuous preference representation and annotation consistency, establishing a foundation for subjective alignment in human-centered dialogue systems.
  arXiv  Detail & Related papers  (2025-12-11T12:04:46Z)
• ICPO: Intrinsic Confidence-Driven Group Relative Preference Optimization for Efficient Reinforcement Learning [17.98065634130798]
  We propose the Intrinsic Confidence-Driven Group Relative Preference Optimization method (ICPO)<n>ICPO calculates a preference advantage score for each response by comparing the relative generation probabilities of multiple responses under the same input prompt.<n>We have discovered that the preference advantage score not only alleviates the issues of coarse-grained rewards and reward noise but also effectively curbs overconfident errors.
  arXiv  Detail & Related papers  (2025-11-26T03:10:15Z)
• RAG-Zeval: Towards Robust and Interpretable Evaluation on RAG Responses through End-to-End Rule-Guided Reasoning [64.46921169261852]
  RAG-Zeval is a novel end-to-end framework that formulates faithfulness and correctness evaluation as a rule-guided reasoning task.<n>Our approach trains evaluators with reinforcement learning, facilitating compact models to generate comprehensive and sound assessments.<n>Experiments demonstrate RAG-Zeval's superior performance, achieving the strongest correlation with human judgments.
  arXiv  Detail & Related papers  (2025-05-28T14:55:33Z)
• T2I-Eval-R1: Reinforcement Learning-Driven Reasoning for Interpretable Text-to-Image Evaluation [60.620408007636016]
  We propose T2I-Eval-R1, a novel reinforcement learning framework that trains open-source MLLMs using only coarse-grained quality scores.<n>Our approach integrates Group Relative Policy Optimization into the instruction-tuning process, enabling models to generate both scalar scores and interpretable reasoning chains.
  arXiv  Detail & Related papers  (2025-05-23T13:44:59Z)
• DAFE: LLM-Based Evaluation Through Dynamic Arbitration for Free-Form Question-Answering [12.879551933541345]
  We propose the Dynamic Arbitration Framework for Evaluation (DAFE) to evaluate large language models.<n>DAFE employs two primary LLM-as-judges and engages a third arbitrator only in cases of disagreements.<n>We show DAFE's ability to provide consistent, scalable, and resource-efficient assessments.
  arXiv  Detail & Related papers  (2025-03-11T15:29:55Z)
• CARMO: Dynamic Criteria Generation for Context-Aware Reward Modelling [27.86204841898399]
  Reward modeling in large language models is susceptible to reward hacking.<n>We propose Context-Aware Reward Modeling (CARMO) to mitigate this problem.<n>We establish a new state-of-the-art performance in zero-shot settings for generative models, achieving a 2.1% improvement on Reward Bench.
  arXiv  Detail & Related papers  (2024-10-28T21:18:49Z)
• Language Model Preference Evaluation with Multiple Weak Evaluators [78.53743237977677]
  GED (Preference Graph Ensemble and Denoise) is a novel approach that leverages multiple model-based evaluators to construct preference graphs.<n>In particular, our method consists of two primary stages: aggregating evaluations into a unified graph and applying a denoising process.<n>We provide theoretical guarantees for our framework, demonstrating its efficacy in recovering the ground truth preference structure.
  arXiv  Detail & Related papers  (2024-10-14T01:57:25Z)
• The Comparative Trap: Pairwise Comparisons Amplifies Biased Preferences of LLM Evaluators [31.520403357740317]
  Large language models (LLMs) are increasingly used as evaluators for natural language generation tasks.<n>LLMs display biased preferences, such as favoring verbosity and authoritative tones.<n>We introduce PRePair, which integrates pointwise reasoning within a pairwise framework.
  arXiv  Detail & Related papers  (2024-06-18T06:43:04Z)
• Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluators [48.54465599914978]
  Large Language Models (LLMs) have demonstrated promising capabilities as automatic evaluators in assessing the quality of generated natural language.<n>LLMs still exhibit biases in evaluation and often struggle to generate coherent evaluations that align with human assessments.<n>We introduce Pairwise-preference Search (PAIRS), an uncertainty-guided search-based rank aggregation method that employs LLMs to conduct pairwise comparisons locally and efficiently ranks candidate texts globally.
  arXiv  Detail & Related papers  (2024-03-25T17:11:28Z)
• Aligning Large Language Models by On-Policy Self-Judgment [49.31895979525054]
  Existing approaches for aligning large language models with human preferences face a trade-off that requires a separate reward model (RM) for on-policy learning. We present a novel alignment framework, SELF-JUDGE, that does on-policy learning and is parameter efficient. We show that the rejecting sampling by itself can improve performance further without an additional evaluator.
  arXiv  Detail & Related papers  (2024-02-17T11:25:26Z)
• DCR-Consistency: Divide-Conquer-Reasoning for Consistency Evaluation and Improvement of Large Language Models [4.953092503184905]
  This work proposes DCR, an automated framework for evaluating and improving the consistency of Large Language Models (LLMs) generated texts. We introduce an automatic metric converter (AMC) that translates the output from DCE into an interpretable numeric score. Our approach also substantially reduces nearly 90% of output inconsistencies, showing promise for effective hallucination mitigation.
  arXiv  Detail & Related papers  (2024-01-04T08:34:16Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.