ResearchBenchmarksImage Classification on ImageNet-1K (Token Accuracy & Efficiency)Image Classification on ImageNet-1K (Token Accuracy & Efficiency)Follow76.87Accuracy (Patch Token)Baseline61.810865.720469.6373.5396Jun 17, 2026Evaluation resultsMethodMethodLinksAccuracy (Patch Token)Accuracy (CLS Token)FLOPs Saving (%)Training Time Saving (%)BaselineBackbone=ViT-S, Teache...Backbone=ViT-S, Teacher model=ViT-G2026.0676.8777.63——LEAPBackbone=ViT-S, Teache...Backbone=ViT-S, Teacher model=ViT-G2026.0676.2277.348.57.1BaselineBackbone=ViT-Tiny, Tea...Backbone=ViT-Tiny, Teacher model=ViT-G2026.0663.3564.4——LEAPBackbone=ViT-Tiny, Tea...Backbone=ViT-Tiny, Teacher model=ViT-G2026.0662.3964.1411.5111.6