Generalized Category Discovery on CIFAR-100 (test)
0.852Accuracy (All)VLM-NCD
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VLM-NCDImage Encoder=CLIP, Knowledge Database=Yes2025.12 | 0.852 | 0.85 | 0.856 | — | — | |
| SIEFormerBase Framework=SimGCD, Backbone=ViT-B/16, Pre-training=DINO2026.02 | 0.833 | 0.844 | 0.812 | — | — | |
| SIEFormerBase Framework=SPTNet, Backbone=ViT-B/16, Pre-training=DINO2026.02 | 0.826 | 0.854 | 0.77 | — | — | |
| CMSBackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.823 | 0.857 | 0.755 | — | — | |
| ProtoGCDBackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.819 | 0.829 | 0.8 | — | — | |
| PromptCALsetting=Inductive GCD, backbone=ViT2022.12 | 0.816 | 0.853 | 0.669 | 0.862 | 0.713 | |
| SPTNetBackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.814 | 0.843 | 0.756 | — | — | |
| PromptCALBackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.812 | 0.842 | 0.753 | — | — | |
| SimGCDBackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.801 | 0.812 | 0.778 | — | — | |
| HypGCDBackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.793 | 0.801 | 0.778 | — | — | |
| ORCA (DINO)setting=Inductive GCD, backbone=ViT2022.12 | 0.777 | 0.836 | 0.539 | 0.836 | 0.666 | |
| SIEFormerBase Framework=GCD, Backbone=ViT-B/16, Pre-training=DINO2026.02 | 0.771 | 0.815 | 0.683 | — | — | |
| DCCLBackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.753 | 0.768 | 0.702 | — | — | |
| XCon2025.12 | 0.742 | 0.812 | 0.603 | — | — | |
| GCD2025.12 | 0.73 | 0.762 | 0.665 | — | — | |
| GCDBackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.73 | 0.762 | 0.665 | — | — | |
| GCDsetting=Inductive GCD, backbone=ViT2022.12 | 0.701 | 0.768 | 0.435 | 0.787 | 0.582 | |
| UNO2025.12 | 0.695 | 0.806 | 0.472 | — | — | |
| UNO+Backbone=ViT-B/16, Pre-training=DINO2026.02 | 0.695 | 0.806 | 0.472 | — | — | |
| ORCABackbone=ViT-B/16, Pre-training=DINO2026.02 | 0.69 | 0.774 | 0.52 | — | — | |
| RankStats2025.12 | 0.582 | 0.776 | 0.193 | — | — | |
| RankStats+Backbone=ViT-B/16, Pre-training=DINO2026.02 | 0.582 | 0.776 | 0.193 | — | — |