---
license: apache-2.0
task_categories:
- multiple-choice
- text-generation
- question-answering
language:
- en
---

This dataset is based on [MMLU](https://arxiv.org/abs/2009.03300), [FLAN](https://arxiv.org/abs/2109.01652), [Big Bench Hard](https://arxiv.org/abs/2210.09261) and [AgiEval English](https://arxiv.org/abs/2304.06364).
The non-"reduced" benchmark is the original benchmark, except for FLAN, which is a sampled version. 
The "reduced" benchmark only contains a few representative tasks in the original ones, such that the performance on the "reduced" benchmark can serve as an approximation to the performance on the original ones.