FLIP

该数据集是评估人工智能推理能力的一个基准，基于Idena区块链上的人体验证任务，主要关注顺序推理、视觉故事讲述和常识。此外，该数据集通过两种四幅图像的排序方式向用户提出挑战，强调了对于健壮的多模态基准的需求。其任务在于识别图像在逻辑上连贯的排序方式。This dataset is a benchmark for evaluating the reasoning capabilities of artificial intelligence, built upon human verification tasks on the Idena blockchain. It primarily focuses on sequential reasoning, visual storytelling, and commonsense knowledge. Additionally, the dataset presents users with challenges via two types of four-image sorting tasks, underscoring the urgent need for robust multimodal benchmarks. The core task of this benchmark is to recognize the logically coherent ordering of the given images.