Clustering on Flowers (NMI, ACC, ARI)
99.8NMI (%)SAGL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SAGL2026.05 | 99.8 | 99.76 | 99.65 | |
| MSRL2026.05 | 99.76 | 99.72 | 99.59 | |
| TURTLE2026.05 | 99.66 | 99.59 | 99.4 | |
| SRLF2026.05 | 98.48 | 94.06 | 95.74 | |
| MIM2026.05 | 96.77 | 97.5 | 95.33 | |
| PRR2026.05 | 93.17 | 94.59 | 90.92 | |
| KEC_TURTLEBackbone=ViT-B/32, Training Strategy=TURTLE2026.04 | 92.4 | 88.4 | 82.7 | |
| TAC_TURTLEBackbone=ViT-B/32, Training Strategy=TURTLE2026.04 | 90.8 | 86.8 | 79.1 | |
| TURTLE (1-space)Backbone=ViT-B/322026.04 | 90.7 | 87.2 | 79.8 | |
| KEC (no train)Backbone=ViT-B/32, Training Strategy=Training-free2026.04 | 87.3 | 72.8 | 67.5 | |
| GradNorm2025.10 | 86.7 | 70.8 | 64.2 | |
| CLIP (k-means)Backbone=ViT-B/32, Training Strategy=Training-free2026.04 | 86.5 | 71.5 | 67.7 | |
| TAC2025.10 | 86 | 66.9 | 58.5 | |
| KEC_TACBackbone=ViT-B/32, Training Strategy=TAC2026.04 | 85.4 | 71.6 | 66.8 | |
| TAC (no train)Backbone=ViT-B/32, Training Strategy=Training-free2026.04 | 84.5 | 69.4 | 64.8 | |
| TACBackbone=ViT-B/32, Training Strategy=TAC2026.04 | 80.4 | 64.2 | 59.6 | |
| CLIP (zero-shot)Backbone=ViT-B/32, Evaluation Protocol=Zero-shot with GT labels2026.04 | 79.4 | 67.5 | 58.7 | |
| SICBackbone=ViT-B/322026.04 | 67.7 | 43.1 | 34 | |
| VLM Caption + ClusterLLM=GPT-4o2026.04 | 60.1 | 41.3 | 30.5 | |
| Neural Tangent Kernel Spectral Clustering2026.02 | 0.883 | 0.694 | 0.61 | |
| GradNorm2026.02 | 0.867 | 0.708 | 0.642 | |
| TAC (KMeans)2026.02 | 0.86 | 0.669 | 0.585 | |
| TAC (SC)2026.02 | 0.86 | 0.67 | 0.595 |