[论文] AVGen-Bench: A Task-Driven Benchmark for Multi-Granular Evaluation of Text-to-Audio-Video Generation

## 论文概要 **研究领域**: AI **作者**: Ziwei Zhou, Zeyuan Lai, Rui Wang **发布时间**: 2025-04-10 **arXiv**: [2504.07073](https://arxiv.org/abs/2504.07073) ## 中文摘要 文本到音视频（T2AV）生成正迅速成为媒体创作的核心接口，但其评估仍然碎片化。现有基准测试大多孤立地评估音频和视频，或依赖粗粒度嵌入相似度，无法捕捉真实提示所需的细粒度联合正确性。我们引入AVGen-Bench，一个用于T2AV生成的任务驱动基准，包含11个真实世界类别的高质量提示。为支持全面评估，我们提出多粒度评估框架，结合轻量级专家模型和多模态大语言模型（MLLM），实现从感知质量到细粒度语义可控性的评估。我们的评估揭示了强视听美学与弱语义可靠性之间的显著差距，包括文本渲染、语音连贯性、物理推理的持续失败，以及音乐音高控制的普遍崩溃。 ## 原文摘要 Text-to-Audio-Video (T2AV) generation is rapidly becoming a core interface for media creation, yet its evaluation remains fragmented. Existing benchmarks largely assess audio and video in isolation or rely on coarse embedding similarity, failing to capture the fine-grained joint correctness required by realistic prompts. We introduce AVGen-Bench, a task-driven benchmark for T2AV generation featuring high-quality prompts across 11 real-world categories. To support comprehensive assessment, we propose a multi-granular evaluation framework that combines lightweight specialist models with Multimodal Large Language Models (MLLMs), enabling evaluation from perceptual quality to fine-grained semantic controllability. Our evaluation reveals a pronounced gap between strong audio-visual aesthetics and weak semantic reliability, including persistent failures in text rendering, speech coherence, physical reasoning, and a universal breakdown in musical pitch control. --- *自动采集于 2025-04-11* #论文 #arXiv #AI #小凯