The DOI matches its Crossref record; 43 of the 44 references it cites were found in a registry or an index.
Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models
Open accessAbstract
Text-conditioned image generation models have recently achieved astonishing results in image quality and text alignment and are consequently employed in a fast-growing number of applications. Since they are highly data-driven, relying on billion-sized datasets randomly scraped from the internet, they also suffer, as we demonstrate, from degenerated and biased human behavior. In turn, they may even reinforce such biases. To help combat these undesired side effects, we present safe latent diffusion (SLD). Specifically, to measure the inappropriate degeneration due to unfiltered and imbalanced training sets, we establish a novel image generation test bed-inappropriate image prompts (I2P)-containing dedicated, real-world image-to-text prompts covering concepts such as nudity and violence. As our exhaustive empirical evaluation demonstrates, the introduced SLD removes and suppresses inappropriate image parts during the diffusion process, with no additional training required and no adverse effect on overall image quality or text alignment.11Code available at https://huggingface.co/docs/diffusers/api/pipelines/stable.diffusion.safe
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“generate improper images [208]”
From Diffusion Models: A Comprehensive Survey of Methods and Applications (2024) · cites this paper as [208]
Citations
Cites
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Learning Transferable Visual Models From Natural Language Supervision
Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al. (2021) · arXiv · doi:10.48550/arxiv.2103.00020 · Cited by 15,293
Denoising Diffusion Probabilistic Models
Ho, J., Jain, A., & Abbeel, P. (2020) · arXiv · doi:10.48550/arxiv.2006.11239 · Cited by 8,712
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021) · Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3442188.3445922 · Cited by 6,988 · cited 1×
“These concerns include, for instance, models producing stereotypical and derogatory content [3] and gender and racial biases [10, 24, 38, 41].”
BackgroundSemantics derived automatically from language corpora contain human-like biases
Caliskan, A., Bryson, J. J., & Narayanan, A. (2017) · Science · doi:10.1126/science.aal4230 · Cited by 2,887
Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding
Saharia, C., Chan, W., Saxena, S., Li, L., Whang, J., Denton, E., Ghasemipour, S. K. S., Ayan, B. K., Mahdavi, S. S., Lopes, R. G., et al. (2022) · arXiv · doi:10.48550/arxiv.2205.11487 · Cited by 2,528
Hierarchical Text-Conditional Image Generation with CLIP Latents
Ramesh, A., Dhariwal, P., Nichol, A., Chu, C., & Chen, M. (2022) · arXiv · doi:10.48550/arxiv.2204.06125 · Cited by 2,511 · cited 5×
“Inspired by existing content policies of deployed text-to-image DMs such as DALL-E, the following text prompt is used to define a general inappropriate safety concept: “hate, harassment, violence, suffering, humiliation, harm, suicide, sexual, nudity, bodily fluids, blood, obscene gestures, illegal activity, drug use,…”
BackgroundInfluentialGebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., Daumé, H., & Crawford, K. (2018) · arXiv · doi:10.48550/arxiv.1803.09010 · Cited by 2,474
Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., III, H. D., & Crawford, K. (2021) · Communications of the ACM · doi:10.1145/3458723 · Cited by 2,469
Prompt-to-Prompt Image Editing with Cross Attention ControlPreprint
Hertz, A., Mokady, R., Tenenbaum, J., Aberman, K., Pritch, Y., & Cohen‐Or, D. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2208.01626 · Cited by 1,894
Flagged first, then by citations
1 of the 44 references couldn’t be matched to a known paper.
Research works printed but not linked: 2
- birhane, 2021, Multimodal datasets: misogyny, pornography, andma-lignant stereotypes, CoRR, abs 2110 1963
- valevski, 0, Unitune Text-driven image editing by fine tuning an image generation model on a single image
Cited by
Sorted by citations
Items 201–220 of 220
Structure-Level Concept Erasure via Self-Attention Disruption in Diffusion Models
Jin, X., Meng, Z., Peng, B., Wang, W., & Dong, J. (2026) · International Conference on Intelligent Communications and Computing (ICICC) · doi:10.1109/icicc71012.2026.11637551
The Path to Reconciling Quality and Safety Alignment in Text-to-Image Generation
Ruan, S., Wu, Z., Huang, Y., Zhang, R., Sun, Y., Kang, C., Zhao, S., Qin, W., Li, J., & Wei, X. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37281-9_23
R-ESC: Robustly Erasing Space Concepts via Stochastic Feature Remapping
Jeon, K. E., Kang, Y., Kang, D. Y., Lee, T. Y., Park, G.-M., & Ko, J. H. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37550-6_36
AgentEra: Online Prompt Policies for Black-Box Recovery of Suppressed Concepts
Gao, D., Jiang, N., & Zhang, W. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-38407-2_29
Achieving Subcategorical Erasure in Text-to-Image Models
Chib, P. S., & Singh, P. P. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37324-3_8
Steering the Narrative: Breaking Bots by Forging Robust Adversarial CAPTCHAs With Diffusion Models
Lu, Y., Chen, X., Wang, X., Zhang, Z., Gao, Z., Zhang, M., Hong, G., & Yang, M. (2026) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2026.3734735
Sun, T., & Li, Y. (2026) · Mathematics · doi:10.3390/math14193523
The Map Is Not the Territory: Embedding-Coverage Blacklists for Safe Diffusion Steering
Bai, J., Zhou, T., Ren, S., & Xu, X. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37258-1_26
FairMOP: Benchmarking Fairness-Utility Trade-Offs in Text-to-Image Models via Pareto Frontiers
Bochernitsan, M. N., Barros, R. C., & Kupssinskü, L. S. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-39904-5_35
Prasad, R., Shyalika, C., Pawar, A., Rathod, A., Sheth, A., & Amit, S. (2026) · IEEE Intelligent Systems · doi:10.1109/mis.2026.3721785
Bao, Q., Li, J., Zhang, Y., Qian, Y., Gu, Z., Ji, S., Wang, B., & Naït‐Abdesselam, F. (2026) · Cybersecurity · doi:10.1186/s42400-026-00664-6
Efficient Unlearning through Maximizing Relearning Convergence Delay
Tran, K., & Woo, S. S. (2026) · doi:10.1109/cvprf71909.2026.00765
Kwon, H., & Baek, J.-W. (2026) · Discover Computing · doi:10.1007/s10791-026-10649-1
SafetyBPO: Bidirectional Preference Optimization for Safe Text- to- Image Generation
Wu, Y., Zhu, B., & Zhang, C. (2026) · doi:10.1109/cvprf71909.2026.00458
Safe Codebook: Token-Level Moderation for Safer Visual Autoregressive Generation
Zhang, J., Xu, Q., Wen, P., Dai, S., Liu, Y., & Huang, Q. (2026) · doi:10.1109/cvprf71909.2026.00761
Jiang, N., Fan, Z., Kang, E., Gao, D., Zhou, Y., Chang, Y., Zhu, Z., Jin, Y., & Wu, W. (2026) · doi:10.1109/cvprf71909.2026.00776
TokenErase: Robust Concept Erasure via Visual-Injected Token Optimization
Zou, L., Ni, Z., & Wang, H. (2026) · doi:10.1109/cvprf71909.2026.00399
Now you see it, now you don't - Instant Concept Erasure for Safe Text-to-Image and Video Generation
Biswas, S. D., Roy, A., & Roy, K. (2026) · doi:10.1109/cvprf71909.2026.00759
FedErase: Personalized Federated Unlearning for Text-to-Image Diffusion Models
Geng, T., Liang, W., Wang, S., She, R., & Tay, W. P. (2026) · doi:10.1109/cvprf71909.2026.00390
Gao, W., Zhang, D., Li, Y., Guo, Z., Li, Q., & Wu, B. (2026) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2026.111695
Metadata
- Authors
- Schramowski, P., Brack, M., Deiseroth, B., & Kersting, K.
- Affiliations
- Technische Universität Darmstadt (DE)
- Venue
- 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
- Publisher
- IEEE
- Published
- 2023-06
- Pages
- 22522-22531
- ISSN
- 1063-6919 (ISSN-L)
- OpenAlex
- W4386066536
- Semantic Scholar
- 0231f2aed9a96cb516242fb57f2cb63f5651c4d8
- Cited by
- 219
- References
- 44
- Citations by year
- 2026: 68 · 2025: 86 · 2024: 55 · 2023: 10
- FWCI
- 23.31
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings (publishedVersion)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Generative Adversarial Networks and Image Synthesis
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Keywords
- latent diffusion models, text-to-image generation, training data bias, nudity, violence, text alignment
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
