OmniGenBench

该数据集是一个全面性的基准测试，旨在评估最先进的大型多模态模型在感知中心任务和认知中心任务中的指令遵循能力。它包含了57个根据特定模型能力细分的多样化子任务，并对主流的文本到图像生成模型进行了评估。任务的重点是评估生成模型在多模态生成任务上的表现。

This dataset is a comprehensive benchmark designed to evaluate the instruction-following capabilities of state-of-the-art large multimodal models in both perception-centric and cognition-centric tasks. It encompasses 57 diverse subtasks segmented according to specific model capabilities, and also conducts evaluations on mainstream text-to-image generation models. The core focus of these tasks lies in assessing the performance of generative models on multimodal generation tasks.