SAEmnesia: Erasing Concepts in Diffusion Models with Supervised Sparse Autoencoders 文章

ArXiv CS.CV2026-06-01NEWSen作者: Enrico Cassano, Riccardo Renzulli, Marco Nurisso, Mirko Zaffaroni, Alan Perotti, Marco Grangetto
查看原文 →
关系图谱

SAEmnesia: Erasing Concepts in Diffusion Models with Supervised Sparse Autoencoders · 相关技术

相关技术

sparse autoencodersequential unlearningconcept erasurefeature centralizationone-to-one concept-neuron mappingsfeature splittingdiffusion modelsconcept unlearningSAEmnesia