AI RESEARCH
DTVI: Dual-Stage Textual and Visual Intervention for Safe Text-to-Image Generation
arXiv CS.CV
•
ArXi:2603.22041v1 Announce Type: new Text-to-Image (T2I) diffusion models have nstrated 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.