Model Overview Card for Meta Llama Guard 3(8B)

Basic Information

• Model Name: Meta Llama Guard 3 (8B)
• Developer/Creator: Meta
• Release Date: July 23, 2023
• Version: 1.0
• Model Type: Language Model for Input/Output Safeguarding

Description

Overview:

Meta Llama Guard 3 (8B) is a language model designed to provide input and output safeguards for human-AI conversations. It focuses on content moderation and safety, ensuring the responses generated by AI systems adhere to predefined safety standards.

Key Features:

• Fine-tuned for content safety classification across multiple languages.
• Capable of classifying prompts and responses to identify safe or unsafe content.
• Implements a safety risk taxonomy for effective moderation.
• Supports zero-shot and few-shot prompting for diverse applications.
• Generates binary decision scores for prompt safety evaluation.

Intended Use:

The model is intended for developers looking to enhance the safety of AI systems, particularly in applications involving conversational agents, customer support bots, and any scenario where user interaction with AI is prevalent.

Language Support:

Meta Llama Guard 3 supports multiple languages, making it suitable for global applications in content moderation.

Technical Details

Architecture:

The model is based on the Llama 3.1 architecture, utilizing an optimized transformer design that incorporates supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) to improve response quality and safety.

Training Data:

Meta Llama Guard 3 was trained on a carefully curated dataset focused on safety risks in AI interactions, ensuring robust performance in identifying harmful content.

• Data Source and Size: The training dataset includes various conversational datasets that emphasize safety and moderation, although specific sizes are not disclosed.
• Knowledge Cutoff: The model's knowledge is current as of March 2023.
• Diversity and Bias: The training data was selected to minimize bias while maximizing the diversity of scenarios encountered in real-world interactions, enhancing the model's robustness.

Performance Metrics and Comparison to Other Models:

Meta-Llama-Guard-3 has shown strong performance metrics:

Usage

Code samples

The model is available on the AI/ML API platform as "Llama Guard 3 (8B)" .

Ethical Guidelines

Meta emphasizes ethical considerations in AI development by promoting transparency regarding the model's capabilities and limitations. The company encourages responsible usage to prevent misuse or harmful applications of generated content.

License Type

The model is licensed for both research and commercial use under an open-source license that promotes ethical AI development while allowing flexibility for various applications.

Get Llama Guard 3 (8B) API here.