1 paper
Aadi Dash, Lennon J. Shikhman, Michael Galarnyk
Lower error from a text-conditioned neural surrogate does not, by itself, show that the model uses the meaning of the text. We examine this attribution problem with OperatorCLIP, c…