Home

Awesome

FlagEval 简体中文


Overview

FlagEval is an open-source evaluation toolkit as well as an open platform for evaluation of large models.

FlagEval aims to cater to three principal evaluation subjects: foundational models, pre-training algorithms, and fine-tuning/compression algorithms. It encompasses four critical evaluation scenarios — Natural Language Processing (NLP), Computer Vision (CV), Audio, and Multimodal, alongside an abundant variety of downstream tasks. You can find more information on our official website flageval.baai.ac.cn.

We're committed to developing scientific, impartial, and clear benchmarks, methodologies, and tools. Our goal is to enable researchers to thoroughly evaluate the effectiveness of foundational models and training algorithms. In addition, we are exploring the use of AI techniques to enhance subjective assessments, increasing both the objectivity and efficiency of our evaluation processes.

FlagEval open-source toolkit now contains follwing sub-projects.

1. mCLIPEval

mCLIPEval is a evaluation toolkit for vision-language models (such as CLIP, Contrastive Language–Image Pre-training).

How to use

Environment Preparation:

Step:

git clone https://github.com/FlagOpen/FlagEval.git
cd FlagEval/mCLIPEval/
pip install -r requirements.txt

Please refer to mCLIPEval/README.md for more details.

2. ImageEval-prompt

ImageEval-prompt is a set of prompts that evaluate text-to-image (T2I) models at a fine-grained level, including entity, style and detail. By conducting comprehensive evaluations at a fine-grained level, researchers can better understand the strengths and limitations of T2I models, in order to further improve their performance.

Please refer to imageEval/README.md for more details.

3. C-SEM

C-SEM innovatively constructs various levels and difficulties of evaluation data to address the potential flaws and inadequacies of current large models. It examines the models' "thinking" process in understanding semantics, referencing human language cognition habits. The currently open-source version, C-SEM v1.0, includes four sub-evaluation items, assessing models' semantic understanding abilities at both the lexical and sentence levels, offering broad applicability for research comparison.

The sub-evaluation items of C-SEM are:

Future iterations of the C-SEM benchmark will continue to evolve, covering more semantic understanding-related knowledge and forming a multi-level semantic understanding evaluation system. Meanwhile, the 【FlagEval large model evaluation platform](https://flageval.baai.ac.cn/#/trending) will integrate the latest versions promptly to enhance the comprehensiveness of evaluating Chinese capabilities of large language models.

Please refer to csem/README.md for more details.

Contact us

License

The majority of FlagEval is licensed under the Apache 2.0 license, however portions of the project are available under separate license terms:

Misc

↳ Stargazers, thank you for your support!

Stargazers repo roster for @FlagOpen/FlagEval

↳ Forkers, thank you for your support!

Forkers repo roster for @FlagOpen/FlagEval

If you find our work helpful, please consider to star🌟 this repo. Thanks for your support!