HUME

Reported on 9 benchmarks across 2 tasks

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision9 results

Image ClusteringonCIFAR-10
ARI
0.776
best: 0.989 (TURTLE (CLIP + DINOv2))
Image ClusteringonCIFAR-10
Accuracy
0.884
best: 0.995 (TURTLE (CLIP + DINOv2))
Image ClusteringonCIFAR-100
ARI
0.377
best: 0.834 (TURTLE (CLIP + DINOv2))
Image ClusteringonCIFAR-100
Accuracy
0.555
best: 0.898 (TURTLE (CLIP + DINOv2))
Image ClusteringonSTL-10
ARI
0.812
best: 0.994 (TURTLE (CLIP + DINOv2))
Image ClusteringonSTL-10
Accuracy
0.908
best: 0.997 (TURTLE (CLIP + DINOv2))
Image ClassificationonSTL-10
Accuracy
90.8
best: 99.7 (TURTLE (CLIP + DINOv2))
Image ClassificationonCIFAR-10
Accuracy
88.4
best: 99.612 (efficient adaptive ensembling)
Image ClassificationonCIFAR-20
Accuracy
55.5
best: 73.2 (MV-MR)