PhyAVBench：面向物理 grounded 文本到音视频生成的音频物理敏感性基准
PhyAVBench: A Challenging Audio Physics-Sensitivity Benchmark for Physically Grounded Text-to-Audio-Video Generation
研究者提出 PhyAVBench，首个系统评估 T2AV、I2AV 和 V2A 模型音频物理 grounding 能力的基准，包含 25.5 小时、11,605 段视频的 PhyAV-Sound-11K 数据集，覆盖 6 个音频物理维度和 41 个细粒度测试点。
Authors:Tianxin Xie, Wentao Lei, Kai Jiang, Guanjie Huang, Pengfei Zhang, Chunhui Zhang, Fengji Ma, Haoyu He, Han Zhang, Jiangshan He, Jinting Wang, Linghan Fang, Lufei Gao, Orkesh Ablet, Peihua Zhang, Ruolin Hu, Shengyu Li, Weilin Lin, Xiaoyang Feng, Xinyue Yang, Yan Rong, Yanyun Wang, Zihang Shao, Zelin Zhao, Chenxing Li, Shan Yang, Wenfu Wang, Meng Yu, Dong Yu, Li Liu
Abstract:Text-to-audio-video (T2AV) generation is central to applications such as filmmaking and world modeling. However, current models often fail to produce physically plausible sounds. Previous benchmarks primarily focus on audio-video temporal synchronization, while largely overlooking explicit evaluation of audio-physics grounding, thereby limiting the study of physically plausible audio-visual generation. To address this issue, we present PhyAVBench, the first benchmark that systematically evaluates the audio-physics grounding capabilities of T2AV, image-to-audio-video (I2AV), and video-to-audio (V2A) models. PhyAVBench offers PhyAV-Sound-11K, a new dataset of 25.5 hours of 11,605 audible videos collected from 184 participants to ensure diversity and avoid data leakage. It contains 337 paired-prompt groups with controlled physical variations that drive sound differences, each grounded with an average of 17 videos and spanning 6 audio-physics dimensions and 41 fine-grained test points. Each prompt pair is annotated with the physical factors underlying their acoustic differences. Importantly, PhyAVBench leverages paired text prompts to evaluate this capability. We term this evaluation paradigm the Audio-Physics Sensitivity Test (APST) and introduce a novel metric, the Contrastive Physical Response Score (CPRS), which quantifies the acoustic consistency between generated videos and their real-world counterparts. We conduct a comprehensive evaluation of 17 state-of-the-art models. Our results reveal that even leading commercial models struggle with fundamental audio-physical phenomena, exposing a critical gap beyond audio-visual synchronization and pointing to future research directions. We hope PhyAVBench will serve as a foundation for advancing physically grounded audio-visual generation. Prompts, ground-truth, and generated video samples are available at this https URL.
| Comments: | 6 major physical dimensions, 41 fine-grained test points, 337 groups of variable-controlled test samples, 11,605 newly recorded videos |
| Subjects: | Sound (cs.SD); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2512.23994 [cs.SD] |
| (or arXiv:2512.23994v5 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2512.23994 arXiv-issued DOI via DataCite |
Submission history
From: Tianxin Xie [view email]
[v1]
Tue, 30 Dec 2025 05:22:31 UTC (3,620 KB)
[v2]
Tue, 7 Apr 2026 11:35:10 UTC (17,164 KB)
[v3]
Wed, 8 Apr 2026 05:21:59 UTC (17,164 KB)
[v4]
Mon, 18 May 2026 09:30:21 UTC (17,164 KB)
[v5]
Sun, 11 Oct 2026 06:54:11 UTC (16,686 KB)
来源：arXiv：cs.AI · arxiv.org