UniSRM:一种用于基于推理的细粒度评估的统一语音奖励模型
UniSRM: A Unified Speech Reward Model for Reasoning-Based Fine-grained Assessment
AI总结:
提出UniSRM统一语音奖励模型,通过两阶段流水线和推理一致性奖励机制,实现多维度、可解释的语音质量评估,在多种任务上优于现有方法。
摘要:
评估语音生成仍然严重依赖人类判断,如平均意见分数(MOS),这些方法昂贵、主观且难以大规模复现。尽管最近一些研究开始探索基于AudioLLM的评判模型,但现有努力通常仅针对狭窄的场景(例如,话语级质量或单轮对话),并且对多样化语音生成任务和评估维度的覆盖有限。在这项工作中,我们提出了UniSRM,一种统一的语音奖励模型,能够支持具有可靠推理的多维、可解释的奖励信号。为了支持训练和评估,我们引入了UniSRM-Data和UniSRM-Bench,涵盖了从话语级质量到上下文级连贯性的语音评估任务。基于该数据集,我们提出了统一的语音奖励模型UniSRM,采用两阶段流水线实现基于推理的细粒度评估。此外,我们引入了推理一致性奖励以提高推理过程的可靠性。实验表明,UniSRM在广泛的语音评估任务中提供了更可靠且与人类一致的判断,为可扩展和统一的语音质量评估提供了实用基础。
英文摘要:
Evaluating speech generation still relies heavily on human judgments, such as Mean Opinion Score (MOS), which are expensive, subjective, and difficult to reproduce at scale. While a few recent studies have begun to explore AudioLLM-based judge models, existing efforts typically target only a narrow set of scenarios (e.g., utterance-level quality or single-turn dialogue) and provide limited coverage of diverse speech generation tasks and evaluation dimensions. In this work, we propose UniSRM, a unified speech reward model that can support multi-dimensional, interpretable reward signals with reliable reasoning. To support training and evaluation, we introduce UniSRM-Data and UniSRM-Bench, covering speech evaluation tasks from utterance-level quality to context-level coherence. Based on this dataset, we present the unified speech reward model, UniSRM, with a two-stage pipeline that enables reasoning-based fine-grained assessment. Furthermore, we introduce Reasoning-Consistent Rewards to improve the reliability of the reasoning process. Experiments show that UniSRM delivers more reliable and human-aligned judgments across a broad range of speech evaluation tasks, offering a practical foundation for scalable and unified evaluation of speech quality.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。