PaperHnS
Hide and Seek
A black-box concept-level attribution method for visual models that quantifies concept contributions by masking target concept regions and observing changes in model output. It requires neither access to model weights nor concept-annotated datasets, and supports image-level local attribution.
Timeline (last 90 days)
Oct 12
Hide and Seek (HnS) is a concept-level attribution method that requires neither access to model weights nor a concept-annotated dataset.
Unverified50%