List your company
ResearchFormerly WizWand

Image Classification on Stanford Cars (Accuracy)

95.22Top-1 AccuracyθB fine-tune
43.927257.243670.5683.8764Mar 20, 2024Aug 4, 2024Dec 20, 2024May 7, 2025Sep 21, 2025Feb 6, 2026Jun 24, 2026
Updated 4mo ago

Evaluation results

MethodLinks
θB fine-tuneTarget Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, Evaluation Protocol=fine-tune2026.0595.22
BiCoSource Backbone=ViT-B/16, Source Pre-training=LAION-2B, Target Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, N=100, Evaluation Protocol=task-vector transfer2026.0593.79
THESEUSSource Backbone=ViT-B/16, Source Pre-training=LAION-2B, Target Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, N=100, Evaluation Protocol=task-vector transfer2026.0593.17
θB + τA paddedSource Backbone=ViT-B/16, Source Pre-training=LAION-2B, Target Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, Evaluation Protocol=task-vector transfer2026.0593.13
θB zero-shotTarget Backbone=ViT-L/14, Target Pre-training=Datacomp-XL, Evaluation Protocol=zero-shot2026.0593.06
Traditional MTLBackbone=ViT-L/142026.0492.84
Auto-SwitchType=Dynamic, Storage (MB)=556.88, Backbone=ViT-L/142026.0492.61
Auto-FlexSwitch+KL (λ = 0.3)Type=Dynamic, Storage (MB)=80.69, Backbone=ViT-L/142026.0492.58
Auto-FlexSwitch+MSE (λ = 0.05)Type=Dynamic, Storage (MB)=98.41, Backbone=ViT-L/142026.0492.56
theta_A fine-tuneN=–2026.0592.5
Auto-FlexSwitch+CKA (λ = 3)Type=Dynamic, Storage (MB)=89.76, Backbone=ViT-L/142026.0492.35
IndividualBackbone=ViT-L/142026.0492.34
FAR-TinyParams (M)=7.5, FLOPs (G)=1.45, Structured Compression=false, Fine-tuning=true2025.0592.3
FAR-BaseParams (M)=83.2, FLOPs (G)=17.31, Structured Compression=false, Fine-tuning=true2025.0592.3
FAR-SmallParams (M)=23.9, FLOPs (G)=4.82, Structured Compression=false, Fine-tuning=true2025.0592.1
DeiT-BaseParams (M)=86.5, FLOPs (G)=17.56, Structured Compression=false, Fine-tuning=true2025.0592.1
θA fine-tuneSource Backbone=ViT-B/16, Source Pre-training=LAION-2B, Evaluation Protocol=fine-tune2026.0592.1
DeiT-SmallParams (M)=22.1, FLOPs (G)=4.60, Structured Compression=false, Fine-tuning=true2025.0591.7
MoW-MergingType=Dynamic, Storage (MB)=5159.51, Backbone=ViT-L/142026.0491.32
EMR-MergingType=Dynamic, Storage (MB)=1633.46, Backbone=ViT-L/142026.0490.71
DeiT-TinyParams (M)=5.7, FLOPs (G)=1.25, Structured Compression=false, Fine-tuning=true2025.0590.5
FAR-Small (Pruned)Params (M)=20.3, FLOPs (G)=3.90, Structured Compression=true, Fine-tuning=true2025.0590.5
AdaMerging++Type=Static, Backbone=ViT-L/142026.0490.48
Twin-MergingType=Dynamic, Storage (MB)=1780.62, Backbone=ViT-L/142026.0490.16
theta_B fine-tuneN=–2026.0590.11
DF-MergeType=Static, Backbone=ViT-L/142026.0489.95
FAR-Base (Pruned)Params (M)=73.0, FLOPs (G)=14.80, Structured Compression=true, Fine-tuning=true2025.0589.8
AdaMergingType=Static, Backbone=ViT-L/142026.0489.65
S2-FracMixBackbone=CLIP ResNet-50, Evaluation Protocol=Fine-tuning2026.0689.63
FAR-Tiny (Pruned)Params (M)=5.7, FLOPs (G)=1.08, Structured Compression=true, Fine-tuning=true2025.0589.5
S2-FracMixBackbone=DINOv2 ViT-S, Evaluation Protocol=Fine-tuning2026.0688.68
CutMixBackbone=DINOv2 ViT-S, Evaluation Protocol=Fine-tuning2026.0688.53
MixupBackbone=DINOv2 ViT-S, Evaluation Protocol=Fine-tuning2026.0688.32
No AugmentationBackbone=DINOv2 ViT-S, Evaluation Protocol=Fine-tuning2026.0688.3
MixupBackbone=CLIP ResNet-50, Evaluation Protocol=Fine-tuning2026.0688.27
CutMixBackbone=CLIP ResNet-50, Evaluation Protocol=Fine-tuning2026.0688.24
RegMeanType=Static, Backbone=ViT-L/142026.0487.17
No AugmentationBackbone=CLIP ResNet-50, Evaluation Protocol=Fine-tuning2026.0687.1
ViT-BAdaptation Strategy=Fine-tuning, Params=85.8M, Pre-training Dataset=ImageNet-1K2026.0586.8
BiCoN=1002026.0585.23
bViT-BAdaptation Strategy=Fine-tuning, Params=7.8M, Pre-training Dataset=ImageNet-1K2026.0585.2
BiCoN=502026.0584.9
BiCoN=202026.0584.6
BiCoN=102026.0584.59
theta_B zero-shotN=–2026.0584.5
THESEUSN=502026.0584.4
THESEUSN=1002026.0584.39
THESEUSN=202026.0584.16
THESEUSN=102026.0584.11
Ties-MergingType=Static, Backbone=ViT-L/142026.0483.09
theta_B + tau_A^paddedN=–2026.0582.83
Task-ArithmeticType=Static, Backbone=ViT-L/142026.0482.14
Fisher MergingType=Static, Backbone=ViT-L/142026.0481.57
Weight-AveragingType=Static, Backbone=ViT-L/142026.0481.54
ViT-BAdaptation Strategy=LoRA, FFN, Params=738K, Pre-training Dataset=ImageNet-1K2026.0581.2
ViT-BAdaptation Strategy=LoRA, QV, Params=296K, Pre-training Dataset=ImageNet-1K2026.0581.1
DAREType=Static, Backbone=ViT-L/142026.0479.41
bViT-BAdaptation Strategy=LoRA, FFN, Params=62K, Pre-training Dataset=ImageNet-1K2026.0579.2
Auto-FlexSwitch+CKA (λ = 3)Type=Dynamic, Storage (MB)=47.26, Backbone=ViT-B/322026.0479.12
Auto-SwitchType=Dynamic, Storage (MB)=185.28, Backbone=ViT-B/322026.0478.47
Auto-FlexSwitch+KL (λ = 0.3)Type=Dynamic, Storage (MB)=41.51, Backbone=ViT-B/322026.0478.4
Auto-FlexSwitch+MSE (λ = 0.05)Type=Dynamic, Storage (MB)=50.72, Backbone=ViT-B/322026.0478.21
Pre-trainedBackbone=ViT-L/142026.0477.86
IndividualBackbone=ViT-B/322026.0477.66
bViT-BAdaptation Strategy=LoRA, QV, Params=25K, Pre-training Dataset=ImageNet-1K2026.0576.8
Traditional MTLBackbone=ViT-B/322026.0474.76
Twin-MergingType=Dynamic, Storage (MB)=711.10, Backbone=ViT-B/322026.0473.26
EMR-MergingType=Dynamic, Storage (MB)=540.19, Backbone=ViT-B/322026.0472.75
AdaMerging++Type=Static, Backbone=ViT-B/322026.0469.89
MoW-MergingType=Dynamic, Storage (MB)=2171.89, Backbone=ViT-B/322026.0468.95
AdaMergingType=Static, Backbone=ViT-B/322026.0468.85
RAR (LLaVA1.5)Shots=162024.0367.8
Fisher MergingType=Static, Backbone=ViT-B/322026.0467.07
RegMeanType=Static, Backbone=ViT-B/322026.0466.87
MMRL + NeRP2026.0566.6
MMA + NeRP2026.0566.53
bViT-BAdaptation Strategy=Time embedding tuning, Params=9K, Pre-training Dataset=ImageNet-1K2026.0566.2
PromptSRC + NeRP2026.0566.2
MMA2026.0566.13
MMRL2026.0566.1
PromptSRC2026.0565.7
CoCoOp + NeRP2026.0565.35
CoCoOp2026.0565.32
DF-MergeType=Static, Backbone=ViT-B/322026.0464.53
Weight-AveragingType=Static, Backbone=ViT-B/322026.0463.34
CLIP+KNNShots=162024.0363.2
Task-ArithmeticType=Static, Backbone=ViT-B/322026.0462.06
DAREType=Static, Backbone=ViT-B/322026.0462.06
Ties-MergingType=Static, Backbone=ViT-B/322026.0461.42
RAR (LLaVA1.5)Shots=82024.0361.2
Pre-trainedBackbone=ViT-B/322026.0459.64
CLIP+KNNShots=82024.0356.1
bViT-BAdaptation Strategy=Linear probing, Params=—, Pre-training Dataset=ImageNet-1K2026.0554.8
RAR (LLaVA1.5)Shots=42024.0354.4
CLIP2026.0452.13
CLIP+KNNShots=42024.0349.2
FARE2026.0448.68
AdvFLYPfull2026.0448.1
ViT-BAdaptation Strategy=Linear probing, Params=—, Pre-training Dataset=ImageNet-1K2026.0548
RAR (LLaVA1.5)Shots=22024.0345.9