$λ$-JEPA Spectral Anti-Collapse Regularization for Self-Supervised Learning

Open paper
Benchmark Model Rank Results
linear-probe-image-classification-on-fgvc-aircraftλ-JEPA ViT-B/16 (linear probe)#3Accuracy: 61.1
linear-probe-image-classification-on-dtdλ-JEPA ViT-B/16 (linear probe)#4Accuracy: 73.2
image-classification-on-flowers-102λ-JEPA ViT-B/16, 100 ep (linear probe)#53Accuracy: 92.2
image-classification-on-cifar-100λ-JEPA ViT-B/16 (linear probe)#88Percentage correct: 85.4
image-classification-on-cifar-10λ-JEPA ViT-B/16 (linear probe)#117Percentage correct: 96.9

Rank counts only results with a code link.