mindware/arc-agi-mega

ARC Mega是一个大型数据集，混合了抽象和推理语料库（ARC）风格的提示和非ARC指令/答案对。数据集分为仅包含ARC部分和非ARC部分，保留了MindsAI格式的CSV文件，并为规范的ARC任务添加了精简的JSON表示。经过过滤，移除了合成数据集助手提示、带有系统/用户/GPT前缀的行以及格式错误的ARC重建。

ARC Mega is a large-scale mixture of Abstraction and Reasoning Corpus (ARC) style prompts and non-ARC instruction/answer pairs. The dataset is split into ARC-only and non-ARC partitions, preserving the MindsAI CSV format and adding a minified JSON representation for canonical ARC tasks. Filtering has removed synthetic dataset-helper prompts, rows with system/user/GPT prefixes, and malformed ARC reconstructions.