Thanks a lot for offering a great dataset.
I have one question about the overfit_type = 'N'.
In the dataset you offered, overfit_type is divided into 4 categories, exact verbatim as MV, retrieval verbatim as RV, template verbatim as TV, and no verbatim as N.
For the overfit_type = 'N', are they can be recognized as false positives? Does it mean there are probably no memorized images for those prompts?
Dear, Ryan.
Thanks a lot for offering a great dataset. I have one question about the overfit_type = 'N'. In the dataset you offered, overfit_type is divided into 4 categories, exact verbatim as MV, retrieval verbatim as RV, template verbatim as TV, and no verbatim as N. For the overfit_type = 'N', are they can be recognized as false positives? Does it mean there are probably no memorized images for those prompts?
Best, Chunsan.