Image Clustering on 20 datasets (average)
58.4NMIKEC_TURTLE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| KEC_TURTLEBackbone=ViT-B/32, Training Strategy=TURTLE2026.04 | 58.4 | 59.3 | 40.9 | |
| KEC (no train)Backbone=ViT-B/32, Training Strategy=Training-free2026.04 | 58 | 56.6 | 38.5 | |
| KEC_TACBackbone=ViT-B/32, Training Strategy=TAC2026.04 | 57.9 | 59 | 40.3 | |
| TAC_TURTLEBackbone=ViT-B/32, Training Strategy=TURTLE2026.04 | 56.2 | 55.8 | 38.1 | |
| TURTLE (1-space)Backbone=ViT-B/322026.04 | 56.1 | 55.5 | 37.8 | |
| CLIP (zero-shot)Backbone=ViT-B/32, Evaluation Protocol=Zero-shot with GT labels2026.04 | 55.9 | 56.7 | 37.5 | |
| CLIP (k-means)Backbone=ViT-B/32, Training Strategy=Training-free2026.04 | 55.6 | 53.6 | 34.8 | |
| TAC (no train)Backbone=ViT-B/32, Training Strategy=Training-free2026.04 | 55.3 | 54.2 | 35.1 | |
| TACBackbone=ViT-B/32, Training Strategy=TAC2026.04 | 55.2 | 55.4 | 37.2 | |
| SICBackbone=ViT-B/322026.04 | 54.9 | 53 | 35.6 | |
| VLM Caption + ClusterLLM=GPT-4o2026.04 | 50.7 | 51.9 | 32.7 |