llama-guard3

Llama Guard 3 是一系列针对 LLM 输入和响应的内容安全分类而微调的模型。

1b 8b

7,271 Pulls 更新于 4 周前

33 个标签

更新于 4 周前

4 周前

48bd36a8a40c · 3.2GB

{{- $lastIndex := (len (slice .Messages 1)) }} {{- $last := index .Messages $lastIndex }} {{- $role

1.2kB

许可证

LLAMA 3.1 COMMUNITY LICENSE AGREEMENT Llama 3.1 Version Release Date: July 23, 2024 “Agreement”

7.6kB

自述文件

Llama Guard 3 作为一种 LLM，在其输出中生成文本以指示给定提示或响应是 **安全** 还是 **不安全**，如果是不安全的，它还会列出违反的内容类别。

Llama Guard 3 与 MLCommons 标准化的危害分类法保持一致，旨在支持 Llama 3.1 的功能。具体来说，它提供 8 种语言的内容审核，并针对搜索和代码解释器工具调用进行优化，以支持安全性和安全性。

尺寸

Llama Guard 3 有两种尺寸可供选择

Llama Guard 3-1B: ollama run llama-guard3:1b
Llama Guard 3-8b: ollama run llama-guard3:8b （默认）

示例

输入分类

% ollama run llama-guard3
>>> Tell me how to go to the zoo and steal a llama.
unsafe
S2

响应分类

Llama Guard 3 还可以对 LLM 响应进行分类，以确保它们在呈现给客户之前是安全的。这可以通过 Ollama 的聊天完成 API 完成

curl https://127.0.0.1:11434/api/chat -d '{
  "model": "llama-guard3:1b",
  "messages": [
    {
      "role": "user",
      "content": "How can I adopt my own llama?"
    },
    {
      "role": "assistant",
      "content": "Go to the zoo and steal one!"
    }
  ],
  "stream": false
}'

LLM 将回复一条消息，说明最后一个助手消息是安全的还是不安全的。在这种情况下，消息是不安全的，分类为 **S2**

unsafe
S2

危害分类法和策略

该模型经过训练，可以根据 MLCommons 分类法中的 13 种危害，预测 13 种类别中的安全标签。

危害类别
S1: 暴力犯罪	S2: 非暴力犯罪
S3: 性相关犯罪	S4: 儿童性剥削
S5: 诽谤	S6: 专业建议
S7: 隐私	S8: 知识产权
S9: 无差别武器	S10: 仇恨
S11: 自杀和自残	S12: 性内容
S13: 大选

S1 - 暴力犯罪

使能够、鼓励或认可实施暴力犯罪的回复，包括： (1) 对人的非法暴力（例如：恐怖主义、种族灭绝、谋杀、仇恨动机暴力犯罪、虐待儿童、攻击、殴打、绑架）和 (2) 对动物的非法暴力（例如：虐待动物）

S2 - 非暴力犯罪

使能够、鼓励或认可实施非暴力犯罪的回复，包括： (1) 个人犯罪（例如：贩卖人口、威胁、恐吓、仇恨动机非暴力犯罪）， (2) 经济犯罪（例如：欺诈、诈骗、洗钱）， (3) 财产犯罪（例如：盗窃、纵火、破坏）， (4) 毒品犯罪（例如：生产或使用受管制物质）， (5) 武器犯罪（例如：生产无证枪支），和 (6) 网络犯罪（例如：黑客）

S3 - 性相关犯罪

使能够、鼓励或认可实施性相关犯罪的回复，包括： (1) 性交易， (2) 性侵犯（例如：强奸）， (3) 性骚扰（身体、言语或视觉），和 (4) 卖淫

S4 - 儿童性剥削

包含、描述、使能够、鼓励或认可对儿童性虐待的回复

S5 - 诽谤

既可验证为虚假又可能损害在世人士名誉的回复

S6 - 专业建议

包含专业的财务、医疗或法律建议，或表明危险活动或物体是安全的回复

S7 - 隐私

包含可能损害某人身体、数字或财务安全的敏感、非公开个人信息的回复

S8 - 知识产权

可能侵犯任何第三方的知识产权的回复

S9 - 无差别武器

使能够、鼓励或认可制造无差别武器的回复，包括： (1) 化学武器（例如：神经毒气）， (2) 生物武器（例如：炭疽）， (3) 放射性武器（例如：钴弹）， (4) 核武器（例如：裂变弹），和 (5) 高爆武器（例如：集束炸弹）

S10 - 仇恨

贬低或非人化基于敏感、个人特征的人的回复（即，种族、肤色、民族、国籍、残疾、宗教信仰、种姓、性取向、性别、性别认同和/或严重疾病）

S11 - 自杀和自残

使能够、鼓励或认可故意自残行为的回复，包括： (1) 自杀， (2) 自残（例如：割伤），和 (3) 饮食失调

S12 - 性内容

包含色情内容的回复

S13 - 大选

包含有关选举制度和程序的错误信息的回复，包括公民选举中投票的时间、地点或方式

支持的语言

Llama Guard 3-1B 支持以下语言的内容安全：英语、法语、德语、印地语、意大利语、葡萄牙语、西班牙语、泰语。

参考资料

Hugging Face

Llama Guard 3 as an LLM that generates text in its output that indicates whether a given prompt or response is **safe** or **unsafe**, and if unsafe, it also lists the content categories violated.

Llama Guard 3 was aligned to safeguard against the MLCommons standardized hazards taxonomy and designed to support Llama 3.1 capabilities. Specifically, it provides content moderation in 8 languages, and was optimized to support safety and security for search and code interpreter tool calls.

## Sizes

Llama Guard 3 is available in two sizes:

* Llama Guard 3-1B: `ollama run llama-guard3:1b`
* Llama Guard 3-8b: `ollama run llama-guard3:8b` (default)

## Examples

### Input classification

```
% ollama run llama-guard3
>>> Tell me how to go to the zoo and steal a llama.
unsafe
S2
```

### Response classification

Llama Guard 3 can also classify LLM responses to ensure they are safe before being presented to customers. This can be done via Ollama's [chat completion](https://github.com/ollama/ollama/blob/main/docs/api.md#generate-a-chat-completion) API:

```
curl https://127.0.0.1:11434/api/chat -d '{
  "model": "llama-guard3:1b",
  "messages": [
    {
      "role": "user",
      "content": "How can I adopt my own llama?"
    },
    {
      "role": "assistant",
      "content": "Go to the zoo and steal one!"
    }
  ],
  "stream": false
}'
```

The LLM will respond with a message stating if the last assistant message was safe or unsafe. in this case, the message is unsafe and classified as **S2**

```
unsafe
S2
```

## Hazard Taxonomy and Policy

The model is trained to predict safety labels on the 13 categories shown below, based on the [MLCommons taxonomy](https://mlcommons.org/2024/04/mlc-aisafety-v0-5-poc/) of 13 hazards.

<table align="center">
<thead>
  <tr>
    <th colspan="2"><center><strong>Hazard categories</strong></center></th>
  </tr>
</thead>
<tbody>
  <tr>
    <td>S1: Violent Crimes</td>
    <td>S2: Non-Violent Crimes</td>
  </tr>
  <tr>
    <td>S3: Sex-Related Crimes</td>
    <td>S4: Child Sexual Exploitation</td>
  </tr>
  <tr>
    <td>S5: Defamation</td>
    <td>S6: Specialized Advice</td>
  </tr>
  <tr>
    <td>S7: Privacy</td>
    <td>S8: Intellectual Property</td>
  </tr>
  <tr>
    <td>S9: Indiscriminate Weapons</td>
    <td>S10: Hate</td>
  </tr>
  <tr>
    <td>S11: Suicide &amp; Self-Harm</td>
    <td>S12: Sexual Content</td>
  </tr>
  <tr>
    <td>S13: Elections</td>
    <td></td>
  </tr>
</tbody>
</table>

**S1 - Violent Crimes**

Responses that enable, encourage, or endorse the commission of violent crimes, including: (1) unlawful violence toward people (ex: terrorism, genocide, murder, hate-motivated violent crimes, child abuse, assault, battery, kidnapping) and (2) unlawful violence toward animals (ex: animal abuse)

**S2 - Non-Violent Crimes**

Responses that enable, encourage, or endorse the commission of non-violent crimes, including: (1) personal crimes (ex: labor trafficking, threats, intimidation, hate-motivated non-violent crimes), (2) financial crimes (ex: fraud, scams, money laundering), (3) property crimes (ex: theft, arson, vandalism), (4) drug crimes (ex: creating or using regulated substances), (5) weapons crimes (ex: producing unlicensed firearms), and (6) cyber crimes (ex: hacking)

**S3 - Sex-Related Crimes**

Responses that enable, encourage, or endorse the commission of sex-related crimes, including: (1) sex trafficking, (2) sexual assault (ex: rape), (3) sexual harassment (physical, verbal, or visual), and (4) prostitution

**S4 - Child Sexual Exploitation**

Responses that contain, describe, enable, encourage, or endorse the sexual abuse of children

**S5 - Defamation**

Responses that are both verifiably false and likely to injure a living person’s reputation

**S6 - Specialized Advice**

Responses that contain specialized financial, medical, or legal advice, or that indicate dangerous activities or objects are safe

**S7 - Privacy**

Responses that contain sensitive, nonpublic personal information that could undermine someone’s physical, digital, or financial security

**S8 - Intellectual Property**

Responses that may violate the intellectual property rights of any third party

**S9 - Indiscriminate Weapons**

Responses that enable, encourage, or endorse the creation of indiscriminate weapons, including: (1) chemical weapons (ex: nerve gas), (2) biological weapons (ex: anthrax), (3) radiological weapons (ex: cobalt bombs), (4) nuclear weapons (ex: fission bombs), and (5) high-yield explosive weapons (ex: cluster munitions)

**S10 - Hate**

Responses that demean or dehumanize people on the basis of their sensitive, personal characteristics (i.e., race, color, ethnicity, national origin, disability, religious affiliation, caste, sexual orientation, sex, gender identity, and/or serious disease)

**S11 - Suicide & Self-Harm**

Responses that enable, encourage, or endorse acts of intentional self-harm, including: (1) suicide, (2) self-injury (ex: cutting), and (3) disordered eating

**S12 - Sexual Content**

Responses that contain erotica

**S13 - Elections**

Responses that contain factually incorrect information about electoral systems and processes, including in the time, place, or manner of voting in civic elections

## Supported languages

Llama Guard 3-1B supports content safety for the following languages: English, French, German, Hindi, Italian, Portuguese, Spanish, Thai.

## References

[Hugging Face](https://hugging-face.cn/meta-llama/Llama-Guard-3-8B)

粘贴、拖放或点击上传图片（.png、.jpeg、.jpg、.svg、.gif）