CLIP_category_singularity

Hypothesis: as the the number of ground truth categories and artificial semantically equivalent categories approaches infinity, the image category labeling capabilities of a Vision Transformer converges to 100%

Dataset labeling tool used in our experiments:

https://github.com/hem9984/Dataset-label

Research Paper:

https://arxiv.org/pdf/2503.12617