AV-Odyssey/AV_Odyssey_Bench

AV-Odyssey是一个评估多模态大型语言模型理解音视频信息能力的综合评估基准，包含26个不同任务和4555个融合文本、视觉和音频元素的问题。数据集为全新收集并由人工注释，特点包括全面的音频属性、广泛的领域以及交织的文本、音频和视觉组件。

AV-Odyssey is a comprehensive evaluation benchmark for assessing the ability of multimodal large language models (MLLMs) to understand audio-visual information, containing 26 different tasks and 4,555 problems integrating text, visual, and audio elements. The dataset is newly collected and annotated by humans, featuring comprehensive audio attributes, extensive domains, and interleaved text, audio, and visual components.