ProGuard: Towards Proactive Multimodal Safeguard

TL;DR

ProGuard is a vision-language model designed to proactively identify and describe out-of-distribution safety risks in multimodal generative models, using a large annotated dataset and reinforcement learning for improved safety moderation.

Contribution

It introduces a large, balanced multimodal safety dataset and a reinforcement learning-based training method for proactive safety risk detection and description.

Findings

ProGuard matches large models in safety classification performance.

It significantly outperforms existing open-source guards in unsafe content categorization.

ProGuard improves OOD risk detection by 52.6% and description by 64.8%.

Abstract

The rapid evolution of generative models has led to a continuous emergence of multimodal safety risks, exposing the limitations of existing defense methods. To address these challenges, we propose ProGuard, a vision-language proactive guard that identifies and describes out-of-distribution (OOD) safety risks without the need for model adjustments required by traditional reactive approaches. We first construct a modality-balanced dataset of 87K samples, each annotated with both binary safety labels and risk categories under a hierarchical multimodal safety taxonomy, effectively mitigating modality bias and ensuring consistent moderation across text, image, and text-image inputs. Based on this dataset, we train our vision-language base model purely through reinforcement learning (RL) to achieve efficient and concise reasoning. To approximate proactive safety scenarios in a controlled setting, we further introduce an OOD safety category inference task and augment the RL objective with a synonym-bank-based similarity reward that encourages the model to generate concise descriptions for unseen unsafe categories. Experimental results show that ProGuard achieves performance comparable to closed-source large models on binary safety classification, substantially outperforms existing open-source guard models on unsafe content categorization. Most notably, ProGuard delivers a strong proactive moderation ability, improving OOD risk detection by 52.6% and OOD risk description by 64.8%.