---
license: llama2
task_categories:
- question-answering
- text2text-generation
- text-generation
language:
- zh
size_categories:
- 1K<n<10K
pretty_name: ANGO
---
ANGO is A Novel Generation-Oriented Chinese LLM evaluation benchmark.

We introduces the format of single-question multiple-keypoints dataset for the first time, which include 171 keypoints accumulated in 4 hierarchical levels and 9 difficulty categories.

The data were exclusively obtained from the Administrative Proficiency Test, which serves as a significant component of the Chinese civil service examination.

We will apply a seasonal system for the leaderboard, updating them every two months. The corresponding test dataset will be announced at the beginning of each season, and some questions will be eliminated at the end of the season.

More details are at our [space](https://huggingface.co/spaces/AngoHF/ANGO-Leaderboard)