Research map / Image generation / Text-to-image diffusion
ProTIP: Probabilistic Robustness Verification on Text-to-Image Diffusion Models against Stochastic Perturbation
Yi Zhang, Yun Tang, Wenjie Ruan, Xiaowei Huang, Siddartha Khastgir, Paul A Jennings, Xingyu Zhao
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 |
|---|---|---|---|
| add sequential analysis with futility and efficacy early stopping rules in the AE test | ProTIP · Single prompt, SD V1.5, 10% perturbation | computational cost (images generated): 246 of 347 Non-AEs identified at stage 1; 97% of Non-AEs detected before stage 5 | Improved |
| use adaptive Hoeffding inequality instead of fixed-size Hoeffding with n=50 (ablation) | ProTIP · Single prompt verification target 0.8 | verification decision correctness: fixed n=50 wrongly asserts fail; ProTIP correct with 78 samples | Improved |
| use adaptive Hoeffding instead of fixed-size Hoeffding with n=200 (ablation) | ProTIP · Single prompt verification target 0.8 | number of perturbations: saves 122 perturbations | Improved |
| use adaptive Hoeffding instead of fixed-size Hoeffding with n=100 (ablation) | ProTIP · Single prompt verification target 0.8 | number of perturbations | No clear effect |
| increase perturbation rate from 10% to 20% | SD V1.5 · Single prompt probabilistic robustness | estimated probabilistic robustness | Hurt |
| use older SD V1.4 instead of SD V1.5 | Stable Diffusion · Single prompt, 10% perturbation | estimated probabilistic robustness | Hurt |
| add misspelling-correction defence (Autocorrect, Pyspellchecker, Gramformer) before the T2I model | Stable Diffusion · Probabilistic robustness under character perturbation | probabilistic robustness | Improved |
| choose Autocorrect over Pyspellchecker and Gramformer as the defence | Stable Diffusion · 36 randomly selected prompts | probabilistic robustness | Improved |
Closest papers by meaning
- Diffusion Models are Certifiably Robust Classifiers (NeurIPS 2024 · similarity 0.90)
- Improved Diffusion-based Generative Model with Better Adversarial Robustness (ICLR 2025 · similarity 0.89)
- Adaptive Median Smoothing: Adversarial Defense for Unlearned Text-to-Image Diffusion Models at Inference Time (ICML 2025 · similarity 0.89)
- CAT: Contrastive Adversarial Training for Evaluating the Robustness of Protective Perturbations in Latent Diffusion Models (ICML 2025 · similarity 0.89)
- Robust Classification via a Single Diffusion Model (ICML 2024 · similarity 0.88)
- Robust Representation Consistency Model via Contrastive Denoising (ICLR 2025 · similarity 0.88)
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.
