Dual-Axis Generative Reward Model Toward Semantic and Turn-taking Robustness in Interactive Spoken Dialogue Models

TL;DR

This paper introduces a Dual-Axis Generative Reward Model for interactive spoken dialogue models, providing reliable, detailed feedback on semantic and timing aspects to improve reinforcement learning-based interaction quality.

Contribution

It presents a novel reward model that offers dual evaluations for semantic and turn-taking quality, addressing the limitations of existing metrics and enabling better RL training for SDMs.

Findings

Achieved state-of-the-art performance on interaction-quality assessment.

Provided reliable, detailed diagnostic feedback for SDMs.

Demonstrated effectiveness across synthetic and real-world datasets.

Abstract

Achieving seamless, human-like interaction remains a key challenge for full-duplex spoken dialogue models (SDMs). Reinforcement learning (RL) has substantially enhanced text- and vision-language models, while well-designed reward signals are crucial for the performance of RL. We consider RL a promising strategy to address the key challenge for SDMs. However, a fundamental barrier persists: prevailing automated metrics for assessing interaction quality rely on superficial proxies, such as behavioral statistics or timing-prediction accuracy, failing to provide reliable reward signals for RL. On the other hand, human evaluations, despite their richness, remain costly, inconsistent, and difficult to scale. We tackle this critical barrier by proposing a Dual-Axis Generative Reward Model, which is trained to understand complex interaction dynamics using a detailed taxonomy and an annotated dataset, produces a single score and, crucially, provides separate evaluations for semantic quality and interaction timing. Such dual outputs furnish precise diagnostic feedback for SDMs and deliver a dependable, instructive reward signal suitable for online reinforcement learning. Our model achieves state-of-the-art performance on interaction-quality assessment across a wide spectrum of datasets, spanning synthetic dialogues and complex real-world interactions.