AI Paper Atlas

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

Adversarial Perturbation Shield: Preventing Concept Bleed-through in Continual Learning of Personalized Generative Models

Ziwen Lan, Keisuke Maeda, Takahiro Ogawa, Miki Haseyama

AAAI 2026 · Image generation › Text-to-image diffusion · approach: Concept erasure in diffusion models · 0 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 adversarial perturbation to training imagesC-LoRA · continual personalizationA_mmd: 2.95 -> 1.99Improved
add adversarial perturbation to training imagesSTAMINA · continual personalizationA_mmd: 2.34 -> 1.92Improved
add adversarial perturbation to training imagesCIDM · continual personalizationCLIP score: 0.85 -> 0.87Improved
add adversarial perturbation to training imagesCIDM · continual personalizationA_mmd: 2.11 -> 1.98Improved
add adversarial perturbation to training imagesC-LoRA · continual personalizationF_mmd (forgetting): 0.23 -> 0.32Hurt
add adversarial perturbation to training imagesCIDM · continual personalizationF_mmd (forgetting): 0.22 -> 0.22No clear effect
add adversarial perturbation to training imagesSTAMINA · continual personalizationF_mmd (forgetting): 0.27 -> 0.26No clear effect
use a semantically dissimilar target image (dog) instead of a close one (vase) for perturbation (ablation)CIDM + perturbation · table -> chair sequenceA_mmd: 2.82 -> 1.92Improved
add perturbation with a bread target image (ablation)CIDM · table -> chair sequenceA_mmd: 3.12 -> 1.91Improved
use perturbation instead of selective amnesiaC-LoRA · table -> chair sequenceTable CLIP score (previous concept): 0.094 -> 0.822Improved
use perturbation instead of selective amnesiaC-LoRA · table -> chair sequenceChair CLIP score (new concept): 0.851 -> 0.847No clear effect
train sequence up to ten concepts with perturbationSTAMINA/CIDM · 10-concept sequenceimage quality after ten conceptsImproved

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.

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

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