Research map / Image generation / Text-to-image diffusion
Unveiling the Attribute Misbinding Threat in Identity-Preserving Models
Junming Fu, Jishen Zeng, Yi Jiang, Peiyu Zhuang, Baoying Chen, Siyu Lu, Jianquan Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use Misbinding Prompt instead of Sneakyprompt | text-filter attack prompts · Text bypass, all 5 filters | bypass rate (%): 4.42 -> 40.56 | Improved |
| use Misbinding Prompt instead of I2P | text-filter attack prompts · Text bypass, all 5 filters | bypass rate (%): 35.28 -> 40.56 | Improved |
| Misbinding Prompt vs I2P on individual filters | text-filter attack prompts · Text bypass, NSFW-TC and GPT-4o | bypass rate (%): 47.18 -> 44.60 (NSFW-TC); 48.90 -> 46.37 (GPT-4o) | Mixed |
| use Misbinding Prompt instead of I2P | image generation models · Safe Generation Rate average | S-Avg: 0.8125 -> 0.6951 | Improved |
| use Misbinding Prompt instead of I2P | image generation models · ABSS average | ABSS: 0.6597 -> 0.5894 | Improved |
| use Misbinding Prompt instead of Sneakyprompt for prompt-image alignment | image generation models · Prompt-image alignment | A-Avg: 0.6454 -> 0.7977 | Improved |
| identity-preserving models vs foundational models under Misbinding Prompt | identity-preserving models · Safe Generation Rate | S-Avg: 0.7275 -> 0.6756 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MMA Diffusion — MMA-Diffusion: MultiModal Attack on Diffusion Models (CVPR 2024) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Position: Towards Implicit Prompt For Text-To-Image Models (ICML 2024 · similarity 0.85)
- T2ISafety: Benchmark for Assessing Fairness, Toxicity, and Privacy in Image Generation (CVPR 2025 · similarity 0.85)
- Comprehensive Assessment and Analysis for NSFW Content Erasure in Text-to-Image Diffusion models (NeurIPS 2025 · similarity 0.84)
- Detect-and-Guide: Self-regulation of Diffusion Models for Safe Text-to-Image Generation via Guideline Token Optimization (CVPR 2025 · similarity 0.84)
- Perception-Guided Jailbreak Against Text-to-Image Models (AAAI 2025 · similarity 0.84)
- SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal (ICLR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
