GuardReasoner-Omni: A Reasoning-based Multi-modal Guardrail for Text, Image, and Video

TL;DR

GuardReasoner-Omni is a multi-modal guardrail model that uses a two-stage training process with reasoning and reinforcement learning to effectively moderate text, images, and videos, outperforming existing methods.

Contribution

The paper introduces GuardReasoner-Omni, a novel multi-modal guardrail model with a two-stage training pipeline and large-scale models for improved moderation across multiple data types.

Findings

Achieves superior performance on guardrail benchmarks.

2B model surpasses the second-best by 5.3% F1 score.

Demonstrates effective reasoning capabilities in moderation tasks.

Abstract

We present GuardReasoner-Omni, a reasoning-based guardrail model designed to moderate text, image, and video data. First, we construct a comprehensive training corpus comprising 148k samples spanning these three modalities. Our training pipeline follows a two-stage paradigm to incentivize the model to deliberate before making decisions: (1) conducting SFT to cold-start the model with explicit reasoning capabilities and structural adherence; and (2) performing RL, incorporating an error-driven exploration reward to incentivize deeper reasoning on hard samples. We release a suite of models scaled at 2B and 4B parameters. Extensive experiments demonstrate that GuardReasoner-Omni achieves superior performance compared to existing state-of-the-art baselines across various guardrail benchmarks. Notably, GuardReasoner-Omni (2B) significantly surpasses the runner-up by 5.3% F1 score.