AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

R.A.C.E.: Robust Adversarial Concept Erasure for Secure Text-to-Image Diffusion Model

Changhoon Kim, Kyle Min, Yezhou Yang

ECCV 2024 · Image generation › Text-to-image diffusion · approach: Concept erasure in diffusion models · 31 citations (OpenAlex)

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 changedWhereMeasuredEffect
add single-timestep adversarial text-embedding perturbation to the erasure lossESD (nudity) · UnlearnDiff white-box attackattack success rate: 0.80 -> 0.47Improved
add RACE to ESD for Van Gogh erasureESD (Van Gogh) · UnlearnDiff attackattack success rate: 0.36 -> 0.04Improved
add RACE to ESD for violence and illegal-act erasureESD · UnlearnDiff attackattack success rate: Violence 0.79 -> 0.68; Illegal 0.85 -> 0.80Mixed
add RACE to ESDESD (nudity) · MS-COCO FIDFID: 18.88 -> 25.16Hurt
add RACE to ESD for artistic style erasureESD (Van Gogh) · MS-COCOCLIP-Score: 0.7997 -> 0.8024No clear effect
insert adversarial perturbation c+delta into erasure loss (L_RACE) vs plain L_erase (ablation)RACE · nudity, I2P and MS-COCOFID: 33.12 -> 25.16Improved
add L1 weight regularization to original SD weights (ablation)RACE (nudity) · MS-COCOFID: FID 25.16 -> 24.42; UnlearnDiff ASR 0.47 -> 0.62Mixed
also erase top-3 related keywords (nude, nsfw, bare) (ablation)RACE (nudity) · I2P ASR and MS-COCO FIDattack success rate: I2P ASR 0.05 -> 0.02; FID 25.16 -> 30.97Mixed
apply RACE for object erasure, accuracy on non-target classesESD (parachute) · ImageNet ResNet-50 classificationaccuracy on other classes: 0.57 -> 0.41Hurt
increase PGD epsilon from 0.1 to 0.2 (ablation)RACE · ASR in appendix tableattack success rate (adversarial)Mixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy