Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Orthogonal Concept Erasure for Diffusion Models

About

Concept erasure has emerged as a promising approach to mitigate undesired or unsafe content in diffusion models, yet existing methods still face significant limitations. While training-based methods are effective, their high computational cost limits scalability. Editing-based methods are more efficient and deployment-friendly, yet they struggle to simultaneously achieve precise concept erasure and preserve overall generative capacity. We identify this core limitation of the editing-based methods as reliance on additive parameter updates. Our empirical analysis reveals that concept semantics primarily depend on neuron direction rather than neuron magnitude, while overall generative capacity relies on the angular geometry of neurons. As additive updates inherently entangle direction, magnitude, and angular geometry, they inevitably introduce unintended interference between concept erasure and overall generation performance. To address this, we propose Orthogonal Concept Erasure (OCE), which reformulates editing-based erasure as multiplicative parameter updates from a geometric perspective. Specifically, OCE applies layer-wise orthogonal transformations derived from a closed-form solution to the parameters, enabling precise concept erasure while preserving the neuron magnitude and angular geometry. Furthermore, to address conflicting constraints in multi-concept erasure, OCE introduces a subspace-level objective with structured subspace manipulation, yielding a more effective and scalable erasure. Extensive experiments on single- and multi-concept erasure demonstrate that OCE outperforms existing methods in concept erasure and non-target preservation, erasing up to 100 concepts in 4.3 s. Code: https://github.com/HansSunY/OCE.

Yuhao Sun, Lingyun Yu, Haoxiang Xu, Fengyuan Miao, Zhuoer Xu, Hongtao Xie• 2026

Related benchmarks

TaskDatasetResultRank
Object ErasureCIFAR-10
Accuracy (Erase)8.79
62
Concept ErasureI2P
I2P Success Rate23
23
Implicit Concept ErasureRing-a-Bell
ASR5
23
Image GenerationMSCOCO
CLIP Score26.33
19
Implicit Concept ErasureMMA
MMA12
14
Text-to-Image GenerationMS-COCO evaluated after 10-concept erasure
CS Score26.53
13
Text-to-Image GenerationMS-COCO evaluated after 50-concept erasure
CLIP Score26.43
13
Text-to-Image GenerationMS-COCO evaluated after 100-concept erasure
CS26.33
13
Concept ErasureRing-a-Bell
ASR43
13
Multi-concept ErasureCelebrity 50 concepts
Acc_e2.66
9
Showing 10 of 15 rows

Other info

Follow for update