What Concepts Lie Within? Detecting and Suppressing Risky Content in Diffusion Transformers

• URL: http://arxiv.org/abs/2605.10180v1
• Date: Mon, 11 May 2026 08:31:57 GMT
• Title: What Concepts Lie Within? Detecting and Suppressing Risky Content in Diffusion Transformers
• Abstract summary: AHV-D&S is a training-free inference-time safeguard for image generation in DiTs.<n>We show that AHV-D&S effectively suppresses sexual, copyrighted-style, and various harmful content while preserving visual quality.
• Score: 41.55824439218019
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The rise of text-to-image (T2I) models has increasingly raised concerns regarding the generation of risky content, such as sexual, violent, and copyright-protected images, highlighting the need for effective safeguards within the models themselves. Although existing methods have been proposed to eliminate risky concepts from T2I models, they are primarily developed for earlier U-Net architectures, leaving the state-of-the-art Diffusion-Transformer-based T2I models inadequately protected. This gap stems from a fundamental architectural shift: Diffusion Transformers (DiTs) entangle semantic injection and visual synthesis via joint attention, which makes it difficult to isolate and erase risky content within the generation. To bridge this gap, we investigate how semantic concepts are represented in DiTs and discover that attention heads exhibit concept-specific sensitivity. This property enables both the detection and suppression of risky content. Building on this discovery, we propose AHV-D\&S, a training-free inference-time safeguard for image generation in DiTs. Specifically, AHV-D\&S quantifies each textual token's sensitivity across all attention heads as an Attention Head Vector (AHV), which serves as a discriminative signature for detecting risky generation tendencies. In the inference stage, we propose a momentum-based strategy to dynamically track token-wise AHVs across denoising steps, and a sensitivity-guided adaptive suppression strategy that suppresses the attention weights of identified risky tokens based on head-specific risk scores. Extensive experiments demonstrate that AHV-D\&S effectively suppresses sexual, copyrighted-style, and various harmful content while preserving visual quality, and further exhibits strong robustness against adversarial prompts and transferability across different DiT-based T2I models.

Related papers

• Unified Safe In-context Image Generation in Multimodal Diffusion Transformers via Restricting Unsafe Information Flows [20.386952794426833]
  Unified Visual Safety Regulator (UVR) is a training-free safe generation framework that regulates unsafe semantics in generated images.<n>UVR mitigates unsafe generation through unified, targeted attention modulation and explicit restriction of harmful information flow.<n>Experiments show that UVR achieves state-of-the-art safety performance by achieving 91% and 77% erase rate in image synthesis and editing tasks.
  arXiv  Detail & Related papers  (2026-06-05T03:43:04Z)
• TINA: Text-Free Inversion Attack for Unlearned Text-to-Image Diffusion Models [92.56562999211809]
  A text-centric paradigm equates erasure with severing the text-to-image mapping, ignoring that the underlying visual knowledge related to undesired concepts still persist.<n>We introduce Tina, a novel Text-free INversion Attack, which enforces this visual-only probe by operating under a null-text condition.<n>Our experiments demonstrate that Tina regenerates erased concepts from models treated with state-of-the-art unlearning.
  arXiv  Detail & Related papers  (2026-03-18T15:25:03Z)
• HERS: Hidden-Pattern Expert Learning for Risk-Specific Vehicle Damage Adaptation in Diffusion Models [0.0]
  Recent advances in text-to-image (T2I) diffusion models have enabled increasingly realistic synthesis of vehicle damage.<n>The ability to generate crash-like imagery challenges the boundary between authentic and synthetic data.<n>We propose HERS, a framework designed to improve fidelity, controllability, and domain alignment of diffusion-generated damage images.
  arXiv  Detail & Related papers  (2026-01-29T10:30:07Z)
• NDM: A Noise-driven Detection and Mitigation Framework against Implicit Sexual Intentions in Text-to-Image Generation [41.058425895887616]
  Text-to-image (T2I) models are vulnerable to generating inappropriate content.<n> implicit sexual prompts, often disguised as seemingly benign terms, can unexpectedly trigger sexual content.<n>We propose NDM, the first noise-driven detection and mitigation framework.
  arXiv  Detail & Related papers  (2025-10-17T15:37:02Z)
• Detect-and-Guide: Self-regulation of Diffusion Models for Safe Text-to-Image Generation via Guideline Token Optimization [22.225141381422873]
  There is a growing concern about text-to-image diffusion models creating harmful content.<n>Post-hoc model intervention techniques, such as concept unlearning and safety guidance, have been developed to mitigate these risks.<n>We propose the safe generation framework Detect-and-Guide (DAG) to perform self-diagnosis and fine-interpret self-regulation.<n>DAG achieves state-of-the-art safe generation performance, balancing harmfulness mitigation and text-following performance on real-world prompts.
  arXiv  Detail & Related papers  (2025-03-19T13:37:52Z)
• TRCE: Towards Reliable Malicious Concept Erasure in Text-to-Image Diffusion Models [53.937498564603054]
  Recent advances in text-to-image diffusion models enable photorealistic image generation, but they also risk producing malicious content, such as NSFW images.<n>To mitigate risk, concept erasure methods are studied to facilitate the model to unlearn specific concepts.<n>We propose TRCE, using a two-stage concept erasure strategy to achieve an effective trade-off between reliable erasure and knowledge preservation.
  arXiv  Detail & Related papers  (2025-03-10T14:37:53Z)
• T2ISafety: Benchmark for Assessing Fairness, Toxicity, and Privacy in Image Generation [39.45602029655288]
  T2ISafety is a safety benchmark that evaluates T2I models across three key domains: toxicity, fairness, and bias.<n>We build a large-scale T2I dataset with 68K manually annotated images and train an evaluator capable of detecting critical risks.<n>We evaluate 12 prominent diffusion models on T2ISafety and reveal several concerns including persistent issues with racial fairness, a tendency to generate toxic content, and significant variation in privacy protection across the models.
  arXiv  Detail & Related papers  (2025-01-22T03:29:43Z)
• SAFREE: Training-Free and Adaptive Guard for Safe Text-to-Image And Video Generation [65.30207993362595]
  Unlearning/editing-based methods for safe generation remove harmful concepts from models but face several challenges.<n>We propose SAFREE, a training-free approach for safe T2I and T2V.<n>We detect a subspace corresponding to a set of toxic concepts in the text embedding space and steer prompt embeddings away from this subspace.
  arXiv  Detail & Related papers  (2024-10-16T17:32:23Z)
• StealthDiffusion: Towards Evading Diffusion Forensic Detection through Diffusion Model [62.25424831998405]
  StealthDiffusion is a framework that modifies AI-generated images into high-quality, imperceptible adversarial examples. It is effective in both white-box and black-box settings, transforming AI-generated images into high-quality adversarial forgeries.
  arXiv  Detail & Related papers  (2024-08-11T01:22:29Z)
• Six-CD: Benchmarking Concept Removals for Benign Text-to-image Diffusion Models [58.74606272936636]
  Text-to-image (T2I) diffusion models have shown exceptional capabilities in generating images that closely correspond to textual prompts.<n>The models could be exploited for malicious purposes, such as generating images with violence or nudity, or creating unauthorized portraits of public figures in inappropriate contexts.<n> concept removal methods have been proposed to modify diffusion models to prevent the generation of malicious and unwanted concepts.
  arXiv  Detail & Related papers  (2024-06-21T03:58:44Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.