About this document

Concept Bottleneck Models: Pang Wei Koh Thao Nguyen Yew Siang Tang Stephen Mussmann Emma Pierson Been Kim Percy Liang by mmmenglish2000 is a document available to read on EtoBox.

The document discusses concept bottleneck models, which allow for the manipulation of high-level concepts in machine learning predictions, enabling better interpretability and human interaction. These models first predict intermediate concepts before making final predictions, allowing users to intervene and correct model mistakes at test time, thus improving accuracy. The authors demonstrate that concept bottleneck models can achieve competitive accuracy compared to standard end-to-end models while providin

Author

mmmenglish2000

Language

EN