Retaining Knowledge and Enhancing Long-Text Representations in CLIP through Dual-Teacher Distillation
Open paper
Benchmark
Model
Rank
Results
image-classification-on-imagenet-v2
LongD-CLIP (ViT-L/14) (extra data)
–
Top 1 Accuracy: 68.8
zero-shot-image-classification-on-imagenet-o
LongD-CLIP (ViT-B/16) (extra data)
–
Accuracy: 41.4
Rank counts only results with a code link.