DTVI: Dual-Stage Textual and Visual Intervention for Safe Text-to-Image Generation

TL;DR

DTVI is a dual-stage inference-time framework that enhances the safety of text-to-image diffusion models by category-aware, sequence-level intervention on prompts and visual outputs, effectively reducing unsafe content.

Contribution

It introduces a novel dual-stage, category-aware intervention method that captures distributed malicious semantics and attenuates unsafe influences during image generation.

Findings

Achieves an average Defense Success Rate of 94.43% on sexual-category benchmarks.

Attains an 88.56% success rate across seven unsafe categories.

Maintains reasonable image quality on benign prompts.

Abstract

Text-to-Image (T2I) diffusion models have demonstrated strong generation ability, but their potential to generate unsafe content raises significant safety concerns. Existing inference-time defense methods typically perform category-agnostic token-level intervention in the text embedding space, which fails to capture malicious semantics distributed across the full token sequence and remains vulnerable to adversarial prompts. In this paper, we propose DTVI, a dual-stage inference-time defense framework for safe T2I generation. Unlike existing methods that intervene on specific token embeddings, our method introduces category-aware sequence-level intervention on the full prompt embedding to better capture distributed malicious semantics, and further attenuates the remaining unsafe influences during the visual generation stage. Experimental results on real-world unsafe prompts, adversarial prompts, and multiple harmful categories show that our method achieves effective and robust defense while preserving reasonable generation quality on benign prompts, obtaining an average Defense Success Rate (DSR) of 94.43% across sexual-category benchmarks and 88.56 across seven unsafe categories, while maintaining generation quality on benign prompts.