List your company
ResearchFormerly WizWand

Image Classification on Food101 (Top-1 Accuracy)

96.38Top-1 AccuracySynCLR + GMAIL
75.132880.648986.16591.6811Dec 9, 2025Dec 27, 2025Jan 14, 2026Feb 2, 2026Feb 20, 2026Mar 10, 2026Mar 29, 2026
Updated 7mo ago

Evaluation results

MethodLinks
SynCLR + GMAILEvaluation Protocol=Zero-shot2026.0296.38
CLIP + GMAILEvaluation Protocol=Zero-shot2026.0295.21
Fine-Tunedbackbone=DINOv3 ViT-B/162025.1293.4
Linear Probebackbone=DINOv3 ViT-B/162025.1293.3
CLIPEvaluation Protocol=Zero-shot2026.0293.08
Task Matrixbackbone=DINOv3 ViT-B/16, best layer=112025.1292.8
SynCLREvaluation Protocol=Zero-shot2026.0291.6
SGDFModel=ViT-L/322026.0390.04
SGDFModel=ViT-B/322026.0389.35
SGDModel=ViT-L/322026.0389.13
SGDModel=ViT-B/322026.0388.56
OURSShots=8, Backbone=ViT-B/16, Templates=72025.1287.8
OURSShots=4, Backbone=ViT-B/16, Templates=72025.1287.6
OURSShots=8, Backbone=ViT-B/16, Templates=12025.1287.3
OURSShots=4, Backbone=ViT-B/16, Templates=12025.1287.1
CLIP-ViT-B/16Layers=-2026.0386.11
OURSShots=1, Backbone=ViT-B/16, Templates=72025.1286.1
OURSShots=1, Backbone=ViT-B/16, Templates=12025.1285.7
LoRA-CLIPShots=1, Backbone=ViT-B/16, Templates=72025.1284.8
LoRA-CLIPShots=1, Backbone=ViT-B/16, Templates=12025.1284.3
CLIP-LoRAShots=8, Backbone=ViT-B/16, Templates=12025.1283.1
CLIP-LoRAShots=8, Backbone=ViT-B/16, Templates=72025.1283.1
LoRA-CLIPShots=4, Backbone=ViT-B/16, Templates=72025.1282.9
LoRA-CLIPShots=4, Backbone=ViT-B/16, Templates=12025.1282.7
SWEETModel Scale=L6H62026.0182.5
WAVEModel Scale=L6H62026.0182.42
WT-SelectModel Scale=L6H62026.0181.95
Col-LnLayers=3, 6, 9, KeepRate=0.72026.0381.29
Iso. PruningModel Scale=L6H62026.0181.27
DMAEModel Scale=L6H62026.0180.78
TCA (Baseline)Layers=3, 6, 9, KeepRate=0.72026.0379.29
Col-LnLayers=0, 3, 6, KeepRate=0.72026.0377.41
TCA (Baseline)Layers=0, 3, 6, KeepRate=0.72026.0375.95