Nice!
An explanation of the APD acronym would be helpful to the readers. In this sense, including this link should be sufficient: https://www.lesswrong.com/posts/EPefYWjuHNcNH4C7E/attribution-based-parameter-decomposition
Nice!
An explanation of the APD acronym would be helpful to the readers. In this sense, including this link should be sufficient: https://www.lesswrong.com/posts/EPefYWjuHNcNH4C7E/attribution-based-parameter-decomposition