BIG-Bench Instruction Induction
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/keirp/automatic_prompt_engineer/tree/main/data/bigbench-ii
下载链接
链接失效反馈官方服务：
资源简介：
该数据集旨在评估跨多种任务的教学引导能力，包括认知推理、逻辑谬误检测、暗示、倒装句、因果判断以及Winowhy等任务。为了提示生成和评估的目的，该数据集被划分为训练集（60%）、验证集（20%）和测试集（20%）。该数据集覆盖了多种任务，并具有较大的数据量，其核心任务是跨多种任务的教学引导。
This dataset is designed to evaluate the instructional guidance capability across diverse tasks, including cognitive reasoning, logical fallacy detection, suggestion, inverted sentences, causal judgment, and Winowhy. For prompt generation and evaluation purposes, this dataset is divided into training set (60%), validation set (20%), and test set (20%). Covering a wide range of tasks with a large dataset volume, its core task focuses on instructional guidance across multiple tasks.
