让自编码器学会可解释的组合特征，能无提示编辑图像
Supervised sparse auto-encoders for interpretable and compositional representations
- 用监督方式训练解码器，学习稀疏概念嵌入与权重
- 在未见组合下仍能重建图像，实现零样本组合泛化
- 适合需要可控语义编辑的生成模型研究者
稀疏自编码器（SAEs）在机制可解释性中重新受到关注，但面临两大挑战：L₁正则项导致非光滑性，影响重建效果与扩展性；以及学习特征与人类语义缺乏对齐。本文通过引入无约束特征模型框架（源自神经坍缩理论）并加以监督，提出仅用解码器的监督式SAE，联合学习稀疏概念嵌入与解码器权重。在Stable Diffusion 3.5上验证，该方法展现出组合泛化能力，能成功重建训练中未出现的概念组合图像，并支持无需修改提示词的特征级语义图像编辑。
原文摘要 · Abstract (English)
Sparse auto-encoders (SAEs) have re-emerged as a prominent method for mechanistic interpretability, yet they face two significant challenges: the non-smoothness of the $L_1$ penalty, which hinders reconstruction and scalability, and a lack of alignment between learned features and human semantics. In this paper, we address these limitations by adapting unconstrained feature models, a mathematical framework from neural collapse theory, and by supervising the task. We supervise (decoder-only) SAEs to reconstruct feature vectors by jointly learning sparse concept embeddings and decoder weights. Validated on Stable Diffusion 3.5, our approach demonstrates compositional generalization, successfully reconstructing images with concept combinations unseen during training, and enabling feature-level intervention for semantic image editing without prompt modification.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。