Spatially Grounded Concept Bottleneck Models via Part-Factorized Attention

Arxiv CS.CV (computer vision)

Paper proposes part-factorized attention to ground concept bottleneck model predictions spatially, improving interpretability on fine-grained recognition. Addresses artifacts from unconstrained attention in existing CBMs.

deep-learninginterpretability

See 1 more source

Hoeffding Concept Bottleneck Models with Applications to Overhead Images8 days agoCl\'ement B\'enard, Manon Arfib, Christophe Labreuche, Victor Qu\'etu

Arxiv CS.CV (computer vision)·Dhanesh Ramachandram

7 days ago