I have a question regarding the prompts for CLIP classifier utilized during the evaluation of gender de-biased models.
Particularly, what prompts did you use for the CLIP classifier to predict the gender of a given image? ("A photo of a female/male" or simply "female/male"?)
Thank you for the great work!
I have a question regarding the prompts for CLIP classifier utilized during the evaluation of gender de-biased models. Particularly, what prompts did you use for the CLIP classifier to predict the gender of a given image? ("A photo of a female/male" or simply "female/male"?)
Thank you :)