arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.03361cs.AI
收藏

ReasonAudio:用于评估文本-音频检索中超越匹配的推理能力的基准

ReasonAudio: A Benchmark for Evaluating Reasoning Beyond Matching in Text-Audio Retrieval

Nanjing University(南京大学);Northwestern Polytechnical University(西北工业大学);Nanyang Technological University(南洋理工大学)

Honglei Zhang, Yuting Chen, Chenpeng Hu, Pengfei Zhou, Siyue Zhang, Yilei Shi

AI总结:

ReasonAudio基准评估文本-音频检索中的否定、时序、共现和时长推理,揭示现有系统性能低下,OmniEmbed-7B得分20.7,远逊于其生成骨干和人类。

摘要:

现有的音频检索基准主要评估语义匹配,缺乏对复杂查询所需的逻辑推理能力的全面评估。我们引入了ReasonAudio,一个用于推理密集型文本-音频检索的基准,评估四种能力:否定、时间顺序、声音共现和声音时长。该基准包含五个合成子任务,涵盖10,000个复合音频片段上的1,000个查询,以及一个自然子任务,涵盖1,000个真实世界片段上的100个查询。对11个最先进的检索系统的评估揭示了显著的局限性:表现最佳的模型OmniEmbed-7B仅达到20.7的总体得分。在一项旨在减少声音事件匹配影响的受控实验中,OmniEmbed-7B的平均准确率为53.8%,而其生成式骨干模型Qwen2.5-Omni-7B-Thinker为70.6%,人类为95.6%。我们的结果凸显了推理密集型音频检索的挑战,并揭示了OmniEmbed-7B与其生成式骨干模型之间的性能差距。

英文摘要:

Existing audio retrieval benchmarks primarily assess semantic matching, lacking comprehensive evaluation of the logical reasoning capabilities required by complex queries. We introduce ReasonAudio, a benchmark for reasoning-intensive Text-Audio Retrieval that evaluates four abilities: negation, temporal order, sound co-occurrence, and sound duration. It comprises five synthetic subtasks with 1,000 queries over 10,000 composite audio clips and a natural subtask with 100 queries over 1,000 real-world clips. Evaluation of 11 state-of-the-art retrieval systems reveals substantial limitations: the best-performing model, OmniEmbed-7B, achieves an overall score of 20.7. In a controlled experiment designed to reduce the influence of sound-event matching, OmniEmbed-7B attains 53.8% average accuracy, compared with 70.6% for its generative backbone, Qwen2.5-Omni-7B-Thinker, and 95.6% for humans. Our results highlight the challenges of reasoning-intensive audio retrieval and reveal a performance gap between OmniEmbed-7B and its generative backbone.

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。

↑