S-Eval

What is S-Eval?

S-Eval is a new comprehensive, multi-dimensional and open-ended safety evaluation benchmark for LLMs consisting of 220,000 evaluation prompts (still in active…

Released

2024-05-23

Evaluates

安全, Safety, LLM, 大语言模型, 安全对齐, Safety Alignment, 合作共建, Co-Built, 不支持, Unsupported

Openness

restricted

Importer

OpenCompass Hub

Reported scores

0

Source provenance

• Original evidence https://hub.opencompass.org.cn/dataset-detail/S-Eval

S-Eval paper, code and dataset

• Paper https://arxiv.org/abs/2405.14191
• Code https://github.com/IS2Lab/S-Eval
• Dataset https://huggingface.co/datasets/IS2Lab/S-Eval

Which sources cite S-Eval?

• S-Eval

Related 安全, Safety, LLM, 大语言模型, 安全对齐, Safety Alignment, 合作共建, Co-Built, 不支持, Unsupported benchmarks

• A-Bench
• A-OKVQA
• AbsPyramid
• ActBench
• ActionAtlas
• Ada-LEval
• AECBench
• AFQMC