LoserLi/MultiTasks

该数据集是一个多任务基准测试集合，包含文本答案和列表答案两种配置。具体包括以下分割：LogicGraph（逻辑图推理）、CVBench（计算机视觉基准测试）、MathVista（数学视觉问题）、OlympiadBench（奥林匹克竞赛问题）、PuzzleVQA（谜题视觉问答）和RefAdv（参考建议）。这些分割覆盖了逻辑推理、视觉理解、数学问题解决和谜题解答等多个领域，旨在评估模型在复杂任务上的性能。数据集结构基于配置组织，每个分割对应不同的数据文件路径。

This dataset is a multi-task benchmark collection that includes two configurations: text_answer and list_answer. It consists of the following splits: LogicGraph (logical graph reasoning), CVBench (computer vision benchmarking), MathVista (mathematical visual problems), OlympiadBench (Olympiad competition problems), PuzzleVQA (puzzle visual question answering), and RefAdv (reference advice). These splits cover various domains such as logical reasoning, visual understanding, mathematical problem-solving, and puzzle solving, aiming to evaluate model performance on complex tasks. The dataset structure is organized based on configurations, with each split corresponding to different data file paths.