Hide and Seek: HIDE Concepts to SEEK explanations for black box models

Key points

• Interpreting machine learning model predictions is a central challenge, particularly for black-box vision models.1
• Concept-based attribution methods interpret models using human-understandable concepts.1
• Given a concept and image set, HnS retrieves concept-aligned images using a vision-language model, localizes the concept within each image, masks the localized region, and measures attribution from the change in the black-box model's output between the original and masked images.1
• We demonstrated the applicability of HnS on diverse tasks, including image captioning and satellite image segmentation, where no existing concept attribution method can be used.1

Sources (1)

• [1]Hide and Seek: HIDE Concepts to SEEK explanations for black box models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 05:56 PM

  We propose Hide and Seek (HnS), a new method for concept-level attribution that does not require access to model weights or labeled concept datasets.

  Interpreting machine learning model predictions is a central challenge, particularly for black-box vision models.

Extractive summary: sentences quoted from the sources.