Llama Guard 3 Vision: Safeguarding Human-AI Image Understanding Conversations

TL;DR

Llama Guard 3 Vision is a multimodal safeguard for human-AI conversations that detects harmful content in both images and text, supporting safer multimodal interactions.

Contribution

It introduces a multimodal safeguard specifically designed for image reasoning in human-AI conversations, extending previous text-only models.

Findings

Strong performance on internal benchmarks

Robustness against adversarial attacks

Effective detection of harmful multimodal prompts

Abstract

We introduce Llama Guard 3 Vision, a multimodal LLM-based safeguard for human-AI conversations that involves image understanding: it can be used to safeguard content for both multimodal LLM inputs (prompt classification) and outputs (response classification). Unlike the previous text-only Llama Guard versions (Inan et al., 2023; Llama Team, 2024b,a), it is specifically designed to support image reasoning use cases and is optimized to detect harmful multimodal (text and image) prompts and text responses to these prompts. Llama Guard 3 Vision is fine-tuned on Llama 3.2-Vision and demonstrates strong performance on the internal benchmarks using the MLCommons taxonomy. We also test its robustness against adversarial attacks. We believe that Llama Guard 3 Vision serves as a good starting point to build more capable and robust content moderation tools for human-AI conversation with multimodal capabilities.