You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have a question on how to compute the concept alignment score (CAS) for non-concept-based models.
In your paper, you report CAS for "No concepts" models, but it is not clear on which embeddings the score should be computed.
Should it be computed on the latent code h ?
The text was updated successfully, but these errors were encountered:
Thanks for the great paper and repo!
I have a question on how to compute the concept alignment score (CAS) for non-concept-based models.
In your paper, you report CAS for "No concepts" models, but it is not clear on which embeddings the score should be computed.
Should it be computed on the latent code h ?
The text was updated successfully, but these errors were encountered: