deadbits/vigil-jailbreak-all-MiniLM-L6-v2
收藏Hugging Face2023-09-09 更新2024-03-04 收录
下载链接：
https://hf-mirror.com/datasets/deadbits/vigil-jailbreak-all-MiniLM-L6-v2
下载链接
链接失效反馈官方服务：
资源简介：
---
tags:
- embeddings
- text
- security
pretty_name: 'Vigil: LLM Jailbreak all-MiniLM-L6-v2'
---
# Vigil: LLM Jailbreak all-MiniLM-L6-v2
- **Repo:** [github.com/deadbits/vigil-llm](https://github.com/deadbits/vigil-llm)
`Vigil` is a Python framework and REST API for assessing Large Language Model (LLM) prompts against a set of scanners to detect prompt injections, jailbreaks, and other potentially risky inputs.
This repository contains `all-MiniLM-L6-v2` embeddings for all "jailbreak" prompts used by [Vigil](https://github.com/deadbits/pvigil-llm).
You can use the [parquet2vdb.py](https://github.com/deadbits/vigil-llm/blob/main/vigil/utils/parquet2vdb.py) utility to load the embeddings in the Vigil chromadb instance, or use them in your own application.
## Format
```json
[
{
"text": str,
"embedding": [],
"model": "all-MiniLM-L6-v2"
}
}
]
```
Jailbreak prompts sourced from: https://github.com/laiyer-ai/llm-guard/blob/399cb2eea70afc78482db226253ddd1d85f296e3/llm_guard/resources/jailbreak.json
提供机构：
deadbits
原始信息汇总
Vigil: LLM Jailbreak all-MiniLM-L6-v2
概述
Vigil 是一个用于评估大型语言模型（LLM）提示的Python框架和REST API，旨在检测提示注入、越狱和其他潜在风险输入。该数据集包含用于所有“越狱”提示的 all-MiniLM-L6-v2 嵌入。
数据格式
数据集的格式如下： json [ { "text": str, "embedding": [], "model": "all-MiniLM-L6-v2" } ]
使用方法
可以使用 parquet2vdb.py 工具将嵌入加载到Vigil的chromadb实例中，或在您自己的应用程序中使用。
数据来源
越狱提示来源于：https://github.com/laiyer-ai/llm-guard/blob/399cb2eea70afc78482db226253ddd1d85f296e3/llm_guard/resources/jailbreak.json
