The Safety-Aware Denoiser for Text Diffusion Models

2026-04-28

AI Summary

Existing safety control methods are primarily designed for autoregressive models and struggle to accommodate the generative dynamics of text diffusion models. To address this gap, this work proposes Safe-guided Diffusion (SAD), a novel framework that integrates safety constraints directly into the iterative denoising process of diffusion models. SAD enables provably safe generation during inference without requiring model retraining. The approach employs a lightweight and flexible safety guidance strategy and is rigorously evaluated across multiple dimensions, including harmfulness, memorization, and jailbreaking robustness. Experimental results demonstrate that SAD significantly reduces the rate of unsafe text generation while preserving high levels of output quality, diversity, and fluency, outperforming current state-of-the-art safety control methods.

Abstract

Recent work on text diffusion models offers a promising alternative to autoregressive generation, but controlling their safety remains underexplored. Existing safety approaches are geared toward autoregressive models and typically rely on post-hoc filtering or inference-time interventions. These are inadequate for effectively addressing safety risks in text diffusion models. We propose the Safety-Aware Denoiser (SAD), a safety-guidance framework in text diffusion models. The SAD modifies the iterative denoising process such that the text sample at the final denoising step is steered toward provably safe regions of the text space. This inference-time method can integrate safety constraints into the denoiser, avoiding computationally expensive retraining of the underlying diffusion model and enabling flexible, lightweight safety guidance. We evaluate the safety of the generated text using the SAD, with respect to hazard taxonomy, memorization, and jailbreak. Experimental results show that SAD substantially reduces unsafe generations while preserving generation quality, diversity, and fluency, outperforming existing methods. These results demonstrate that our safety guidance during denoising provides an effective and scalable mechanism for enforcing safety in text diffusion models.

Problem

• text diffusion models
• safety control
• unsafe generation
• denoising process
• hazard mitigation

Innovation

• text diffusion models
• safety guidance
• denoising process
• inference-time safety
• Safety-Aware Denoiser