Image Classification on CLIP 20-task suite
93.1Average Top-1 AccuracyFine-tuned
Evaluation Results
| Method | Links | |
|---|---|---|
| Fine-tunedBackbone=CLIP-ViT-L/142026.05 | 93.1 | |
| Fine-tunedBackbone=CLIP-ViT-B/322026.05 | 90.3 | |
| TIES-Merging w/ E-PMQBackbone=CLIP-ViT-L/142026.05 | 77.5 | |
| Task Arithmetic w/ E-PMQBackbone=CLIP-ViT-L/142026.05 | 76.7 | |
| TIES-Merging w/ E-PMQBackbone=CLIP-ViT-B/322026.05 | 67.8 | |
| Pre-trainedBackbone=CLIP-ViT-L/142026.05 | 65.6 | |
| Task Arithmetic w/ E-PMQBackbone=CLIP-ViT-B/322026.05 | 64.2 | |
| TIES-MergingBackbone=CLIP-ViT-L/142026.05 | 63 | |
| TIES-Merging w/ GPTQBackbone=CLIP-ViT-L/142026.05 | 61.1 | |
| TIES-Merging w/ RTNBackbone=CLIP-ViT-L/142026.05 | 60.3 | |
| TIES-Merging w/ AWQBackbone=CLIP-ViT-L/142026.05 | 58 | |
| Task ArithmeticBackbone=CLIP-ViT-L/142026.05 | 57.2 | |
| Pre-trainedBackbone=CLIP-ViT-B/322026.05 | 55.6 | |
| TIES-MergingBackbone=CLIP-ViT-B/322026.05 | 55.6 | |
| TIES-Merging w/ GPTQBackbone=CLIP-ViT-B/322026.05 | 53.1 | |
| TIES-Merging w/ AWQBackbone=CLIP-ViT-B/322026.05 | 51.4 | |
| TIES-Merging w/ RTNBackbone=CLIP-ViT-B/322026.05 | 51 | |
| Task ArithmeticBackbone=CLIP-ViT-B/322026.05 | 36.3 | |
| Task Arithmetic w/ GPTQBackbone=CLIP-ViT-B/322026.05 | 35 | |
| Task Arithmetic w/ GPTQBackbone=CLIP-ViT-L/142026.05 | 34.8 | |
| Task Arithmetic w/ AWQBackbone=CLIP-ViT-B/322026.05 | 34 | |
| Task Arithmetic w/ RTNBackbone=CLIP-ViT-L/142026.05 | 33.7 | |
| Task Arithmetic w/ RTNBackbone=CLIP-ViT-B/322026.05 | 33.1 | |
| Task Arithmetic w/ AWQBackbone=CLIP-ViT-L/142026.05 | 27.7 |