Generalized Category Discovery on Fine-grained Avg
76.6Overall AccuracySelEx + AL-GCD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SelEx + AL-GCDBackbone=CLIP ViT-B/16, Category Count Knowledge=Known2026.03 | 76.6 | 78.9 | 75.6 | |
| GETBackbone=CLIP ViT-B/16, Category Count Knowledge=Known2026.03 | 71.5 | 74.8 | 69.8 | |
| SimGCD + AL-GCDBackbone=CLIP ViT-B/16, Category Count Knowledge=Known2026.03 | 70.5 | 75.8 | 68 | |
| CMS + AL-GCDBackbone=CLIP ViT-B/16, Category Count Knowledge=Unknown2026.03 | 69.6 | 76.2 | 66.4 | |
| CMS + AL-GCDBackbone=CLIP ViT-B/16, Clustering Protocol=without the GT number of categories K2026.03 | 69.6 | 76.2 | 66.4 | |
| CMS + AL-GCDBackbone=CLIP ViT-B/16, Category Count Knowledge=Known2026.03 | 69.5 | 75.3 | 66.7 | |
| SelExBackbone=CLIP ViT-B/16, Category Count Knowledge=Known2026.03 | 66.6 | 65.3 | 67.2 | |
| CPTBackbone=CLIP ViT-B/16, Category Count Knowledge=Known2026.03 | 65.9 | 71.3 | 63.2 | |
| CMSBackbone=CLIP ViT-B/16, Category Count Knowledge=Known2026.03 | 64.7 | 74.5 | 59.9 | |
| CMSBackbone=CLIP ViT-B/16, Category Count Knowledge=Unknown2026.03 | 64.5 | 71.4 | 61 | |
| CMSBackbone=CLIP ViT-B/16, Clustering Protocol=without the GT number of categories K2026.03 | 64.5 | 71.4 | 61 | |
| SimGCDBackbone=CLIP ViT-B/16, Category Count Knowledge=Known2026.03 | 64.2 | 71.9 | 60.4 |