paper-with-me

Papers

Post-Hoc Concept Disentanglement: From Correlated to Isolated Concept Representations

2025-03-07 · Eren Erogullari, Sebastian Lapuschkin, Wojciech Samek, Frederik Pahde

Concept Activation Vectors (CAVs) are widely used to model human-understandable concepts as directions within the latent space of neural networks. They are trained by identifying directions from the activations of concept samples to those of non-concept samples. However, this method often produces similar, non-orthogonal directions for correlated concepts, such as "beard" and "necktie" within the CelebA dataset, which frequently co-occur in images of men. This entanglement complicates the interpretation of concepts in isolation and can lead to undesired effects in CAV applications, such as activation steering. To address this issue, we introduce a post-hoc concept disentanglement method that employs a non-orthogonality loss, facilitating the identification of orthogonal concept directions while preserving directional correctness. We evaluate our approach with real-world and controlled correlated concepts in CelebA and a synthetic FunnyBirds dataset with VGG16 and ResNet18 architectures. We further demonstrate the superiority of orthogonalized concept representations in activation steering tasks, allowing (1) the insertion of isolated concepts into input images through generative models and (2) the removal of concepts for effective shortcut suppression with reduced impact on correlated concepts in comparison to baseline CAVs.

📄 PDF Abstract BibTeX arXiv:2503.05522

Code (1)

erenerogullari/cav-disentanglement 공식 구현 pytorch

Tasks

Disentanglement

Similar Papers 제목 키워드 기반

Graph Concept Bottleneck Models

2025-08-19 · Haotian Xu, Tsui-Wei Weng, Lam M. Nguyen, Tengfei Ma arxiv

Concept Bottleneck Models (CBMs) provide explicit interpretations for deep neural networks through concepts and allow intervention with concepts to adjust final predictions. Existing CBMs assume concepts are conditionall…

Image Classification

Scaling-up Disentanglement for Image Translation

2021-03-25 · ICCV 2021 10 · Aviv Gabbay, Yedid Hoshen

Image translation methods typically aim to manipulate a set of labeled attributes (given as supervision at training time e.g. domain label) while leaving the unlabeled attributes intact. Current methods achieve either: (…

DisentanglementDiversityTranslation

ConceptPrism: Concept Disentanglement in Personalized Diffusion Models via Residual Token Optimization

2026-02-23 · Minseo Kim, Minchan Kwon, Dongyeun Lee, Yunho Jeon 외 arxiv

Personalized text-to-image (T2I) generation has emerged as a key application for creating user-specific concepts from a few reference images. The core challenge is concept disentanglement: separating the target concept f…

AttenCraft: Attention-guided Disentanglement of Multiple Concepts for Text-to-Image Customization

2024-05-28 · Junjie Shentu, Matthew Watson, Noura Al Moubayed

With the unprecedented performance being achieved by text-to-image (T2I) diffusion models, T2I customization further empowers users to tailor the diffusion model to new concepts absent in the pre-training dataset, termed…

Disentanglement

Hierarchical Semantic Tree Concept Whitening for Interpretable Image Classification

2023-07-10 · Haixing Dai, Lu Zhang, Lin Zhao, Zihao Wu 외

With the popularity of deep neural networks (DNNs), model interpretability is becoming a critical concern. Many approaches have been developed to tackle the problem through post-hoc analysis, such as explaining how predi…

ClassificationDisentanglementimage-classificationImage Classification