跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.24220 · v1

Erased but Recoverable: On the Fragility of Concept Erasure Under Fine-Tuning

首次提交 2026-11-18 18:56
版本更新 2026-11-18 18:56
雷达观测 2026-11-19 15:04
arXiv ↗PDF ↗

摘要

Existing concept erasure methods for text-to-image diffusion models primarily focus on suppressing target concepts while preserving non-target generation quality. However, in practical applications, open-source diffusion models are often further adapted through DreamBooth, LoRA, or Custom Diffusion. Such adaptations can reshape internal representations, thereby weakening or even reversing the erasure effects achieved by existing methods. We identify erasure persistence as a key yet underexplored dimension of concept erasure and analyze how downstream fine-tuning reshapes internal representations. Our analysis reveals that fine-tuning of the original model does not uniformly modify the model, but preserves certain high-contribution structures, particularly attention heads with stable output subspaces and response magnitudes. We also observe that, after downstream fine-tuning of concept-erased models, these heads' output representations more readily realign with their pretrained counterparts. Motivated by this observation, we propose Stable-Head-Aware Concept Erasure (SHACE), which improves erasure persistence by preserving stable attention structures during concept erasure. SHACE identifies stability-critical attention heads through Attention-Head Stability Profiling and keeps their corresponding parameter blocks unchanged during erasure via Stable-Head Gradient Preservation. Extensive experiments show that SHACE improves resistance to concept recovery across diverse fine-tuning settings while maintaining strong suppression and non-target generation quality.

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-18 18:56 · 最新观测版本

Erased but Recoverable: On the Fragility of Concept Erasure Under Fine-Tuning

Existing concept erasure methods for text-to-image diffusion models primarily focus on suppressing target concepts while preserving non-target generation quality. However, in practical applications, open-source diffusion models are often further adapted through DreamBooth, LoRA, or Custom Diffusion. Such adaptations can reshape internal representations, thereby weakening or even reversing the erasure effects achieved by existing methods. We identify erasure persistence as a key yet underexplored dimension of concept erasure and analyze how downstream fine-tuning reshapes internal representations. Our analysis reveals that fine-tuning of the original model does not uniformly modify the model, but preserves certain high-contribution structures, particularly attention heads with stable output subspaces and response magnitudes. We also observe that, after downstream fine-tuning of concept-erased models, these heads' output representations more readily realign with their pretrained counterparts. Motivated by this observation, we propose Stable-Head-Aware Concept Erasure (SHACE), which improves erasure persistence by preserving stable attention structures during concept erasure. SHACE identifies stability-critical attention heads through Attention-Head Stability Profiling and keeps their corresponding parameter blocks unchanged during erasure via Stable-Head Gradient Preservation. Extensive experiments show that SHACE improves resistance to concept recovery across diverse fine-tuning settings while maintaining strong suppression and non-target generation quality.

查看此版本解读 →查看此版本原文 ↗