AI Paper Atlas

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

EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoders

Xinghao Wang, Dong Li, Wei Yu, Yingwei Pan, Tao Gong, Qi Chu, Nenghai Yu, Ting Yao

ECCV 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 L_cpt and L_leak to L_ctx (ablation)PConvSAE training · HunyuanVideo nudity erasurenudity rate: 24.5 -> 8.09Improved
add temporal consistency loss L_temp (ablation)PConvSAE training · HunyuanVideo nudity erasureSSIM: 47.37 -> 65.69Improved
add L_temp (ablation)PConvSAE training · HunyuanVideo nudity erasurenudity rate: 8.09 -> 7.13Improved
use convolutional single-branch PConvSAE instead of linear SAE (ablation)EraseSAE · HunyuanVideo nudity erasurenudity rate: 12.76 -> 9.06Improved
use convolutional single-branch PConvSAE instead of linear SAE (ablation)EraseSAE · HunyuanVideo nudity erasureSSIM: 56.97 -> 61.37Improved
add dual-branch context/concept partition (ablation)PConvSAE · HunyuanVideo nudity erasurenudity rate: 9.06 -> 7.13Improved
add dual-branch context/concept partition (ablation)PConvSAE · HunyuanVideo nudity erasureSSIM: 61.37 -> 65.69Improved
use spatially-modulated CFG instead of feature subtraction (ablation)EraseSAE · HunyuanVideo nudity erasurenudity rate: 51.30 -> 7.13Improved
use spatially-modulated CFG instead of zeroing masked activations (ablation)EraseSAE · HunyuanVideo nudity erasureSSIM: 53.51 -> 65.69Improved
use spatially-modulated CFG instead of zeroing masked activations (ablation)EraseSAE · HunyuanVideo nudity erasurenudity rate: 6.56 -> 7.13Mixed
use PConvSAE instead of linear SAE (ablation)kernel purity · 5 celebrity identitiespurity: 0.37 -> 0.76Improved

Earlier methods this paper compares with

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