AI Paper Atlas

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

Roots Beneath the Cut: Uncovering the Risk of Concept Revival in Pruning-Based Unlearning for Diffusion Models

Ci Zhang, Zhaojun Ding, Chence Yang, Jun Liu, Xiaoming Zhai, Shaoyi Huang, Beiwen Li, Xiaolong Ma, Jin Lu, Geng Yuan

CVPR 2026 · Image generation › Text-to-image diffusion · approach: Concept erasure in diffusion models · 1 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
apply NMS revival attack to pruned modelConceptPrune SD-v1.5 · ImageNet 12 erased classeserased-class top-1 accuracy: 0.08 -> 0.54Improved
use Quant Recover 4-bit quantization attackConceptPrune SD-v1.5 · ImageNet 12 erased classeserased-class top-1 accuracy: 0.08 -> 0.12No clear effect
add Neuron-Max Scaling instead of neuron-average magnitudes (ablation)NMS attack · ImageNet 12 erased classeserased-class top-1 accuracy: 0.16 -> 0.54Improved
add Top-K sign retention (K=0.8 vs 1.0) (ablation)NMS · golf ball revivalerased-class accuracy: 0.46 -> 0.75Improved
keep too few signs (K=0.2 vs 0.8) (ablation)NMS · golf ball revivalerased-class accuracy: 0.75 -> 0.34Hurt
add Top-0.6 sign retention (vs K=1.0) (ablation)NMS · parachute revivalerased-class accuracy: 0.29 -> 0.71Improved
add Top-0.6 sign retention (vs K=1.0) (ablation)NMS · church revivalerased-class accuracy: 0.11 -> 0.62Improved
add Top-0.4 sign retention (vs K=1.0) (ablation)NMS · gas pump revivalerased-class accuracy: 0.01 -> 0.37Improved
apply NMS attack to Scissorhands-unlearned modelScissorhands · ImageNet 12 erased classeserased-class accuracy: 0.21 -> 0.66Improved
apply NMS attack to SalUn-unlearned modelSalUn · ImageNet 12 erased classeserased-class accuracy: 0.21 -> 0.68Improved
apply NMS attack to ConceptPrune modelConceptPrune · I2P nudity detectionsdetections: 74 -> 118Improved
apply NMS attack to ConceptPrune modelConceptPrune · MMA nudity detectionsdetections: 57 -> 172Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 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