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

arXiv: https://arxiv.org/abs/2605.10180 · HTML: https://mlsanity.com/paper/2605.10180/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Robustness, privacy and adversarial aspects, one of 14 directions in Diffusion models.
- Builds on 22 of its 40 references that are in the map.
- First posted on arXiv 11 May 2026.

## Placed in

- Diffusion models / [Robustness, privacy and adversarial aspects](https://mlsanity.com/diffusion/robustness-privacy-and-adversarial-aspects/)
- LLM safety and security / [Vision-language and text-to-image safety](https://mlsanity.com/safety/vision-language-and-text-to-image-safety/)

## Builds on (most cited first, 22 of 22)

- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [Diffusion Models Beat GANs on Image Synthesis](https://mlsanity.com/paper/2105.05233/) (2021)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [Prompt-to-Prompt Image Editing with Cross Attention Control](https://mlsanity.com/paper/2208.01626/) (2022)
- [Qwen-Image Technical Report](https://mlsanity.com/paper/2508.02324/) (2025)
- [Emerging Properties in Unified Multimodal Pretraining](https://mlsanity.com/paper/2505.14683/) (2025)
- [Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models](https://mlsanity.com/paper/2211.05105/) (2022)
- [Unified Concept Editing in Diffusion Models](https://mlsanity.com/paper/2308.14761/) (2023)
- [MACE: Mass Concept Erasure in Diffusion Models](https://mlsanity.com/paper/2403.06135/) (2024)
- [Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer](https://mlsanity.com/paper/2511.22699/) (2025)
- [Ring-A-Bell! How Reliable are Concept Removal Methods for Diffusion Models?](https://mlsanity.com/paper/2310.10012/) (2023)
- [Show-o2: Improved Native Unified Multimodal Models](https://mlsanity.com/paper/2506.15564/) (2025)
- [To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Now](https://mlsanity.com/paper/2310.11868/) (2023)
- [MMA-Diffusion: MultiModal Attack on Diffusion Models](https://mlsanity.com/paper/2311.17516/) (2023)
- [SAFREE: Training-Free and Adaptive Guard for Safe Text-to-Image And Video Generation](https://mlsanity.com/paper/2410.12761/) (2024)
- [ConceptAttention: Diffusion Transformers Learn Highly Interpretable Features](https://mlsanity.com/paper/2502.04320/) (2025)
- [SPEED: Scalable, Precise, and Efficient Concept Erasure for Diffusion Models](https://mlsanity.com/paper/2503.07392/) (2025)
- [TraSCE: Trajectory Steering for Concept Erasure](https://mlsanity.com/paper/2412.07658/) (2024)
- [T2I-RiskyPrompt: A Benchmark for Safety Evaluation, Attack, and Defense on Text-to-Image Model](https://mlsanity.com/paper/2510.22300/) (2025)
- [Reason2Attack: Jailbreaking Text-to-Image Models via LLM Reasoning](https://mlsanity.com/paper/2503.17987/) (2025)
- [SafeGuider: Robust and Practical Content Safety Control for Text-to-Image Models](https://mlsanity.com/paper/2510.05173/) (2025)

## Built on by (most cited first, 1 of 1)

- [Robust and Generalizable Safety Steering for Text-to-Image Diffusion Transformers](https://mlsanity.com/paper/2605.30049/) (2026)

## Abstract (arXiv)

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.