List your company
ResearchFormerly WizWand

Image Classification on Caltech101 (Accuracy)

96.92AccuracyZS
93.571294.440695.3196.1794Dec 16, 2024Mar 3, 2025May 20, 2025Aug 5, 2025Oct 22, 2025Jan 7, 2026Mar 26, 2026
Updated 7mo ago

Evaluation results

MethodLinks
ZSModel=MLLM, Method=ZS2026.0196.92
CLIP-1+DINO(v2+v3)Setting=Inductive, Test-Time Adaptation (TTA)=false2025.0696.9
COSMIC*Setting=Inductive, Test-Time Adaptation (TTA)=true2025.0696.8
Skip TuningShot=16, Time (s)=962, Mem. (M)=5282024.1296.77
MMFTModel=ViT, Method=Ours2026.0196.73
FLYPModel=ViT, Method=FLYP2026.0196.69
CLIP-1+DINOv2Setting=Inductive, Test-Time Adaptation (TTA)=false2025.0696.6
TOGAShots=16, Venue=-2026.0396.3
FTModel=ViT, Method=FT2026.0196.11
ProSRCShot=16, Time (s)=14403, Mem. (M)=33732024.1296.07
MaPLeShot=16, Time (s)=14070, Mem. (M)=30612024.1296
Skip TuningShot=8, Time (s)=508, Mem. (M)=5282024.1295.9
CLIP-MoE + MoE-GRPOTraining shots=16-shot, Source dataset=ImageNet, Training epochs=32026.0395.9
StatABackbone=ViT-L/14, Scenario=Medium, Batch Size=1282025.0195.8
StatABackbone=ViT-L/14, Scenario=Separate, Batch Size=1282025.0195.8
StatABackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=true, Scenario=Medium2025.0195.8
StatABackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=true, Scenario=Separate2025.0195.8
TOGAShots=8, Venue=-2026.0395.8
CLIP-MoETraining shots=16-shot, Source dataset=ImageNet, Training epochs=32026.0395.8
StatABackbone=ViT-L/14, Scenario=High, Batch Size=1282025.0195.7
StatABackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=true, Scenario=High2025.0195.7
TOGAShots=4, Venue=-2026.0395.7
ProSRCShot=8, Time (s)=7636, Mem. (M)=33732024.1295.67
StatABackbone=ViT-L/14, Scenario=Low, Batch Size=1282025.0195.6
StatABackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=true, Scenario=Low2025.0195.6
Skip TuningShot=4, Time (s)=280, Mem. (M)=5282024.1295.6
CoOpShot=16, Time (s)=3879, Mem. (M)=45512024.1295.57
MaPLeZero-shot=true, Source=ImageNet2026.0295.53
CoFTmode=unsupervised fine-tuning2026.0295.5
OursAnnotation Budget (B%)=5%, Backbone=CLIP ViT-B/16, Protocol=Full Framework (Joint Training)2026.0295.4
DMN*Setting=Inductive, Test-Time Adaptation (TTA)=true2025.0695.4
ProSRCShot=4, Time (s)=3841, Mem. (M)=33732024.1295.27
CLIPBackbone=ViT-L/14, Setting=Zero-shot, Batch Size=1282025.0195.2
CLIPBackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=false2025.0195.2
MaPLeShot=8, Time (s)=7060, Mem. (M)=30612024.1295.2
CLIP-1+DINOv3Setting=Inductive, Test-Time Adaptation (TTA)=false2025.0695.2
MMRL + AmPLevariant=standard2025.1295.17
CoCoOpShot=16, Time (s)=10694, Mem. (M)=28882024.1295.16
ManiPTZero-shot=true, Source=ImageNet2026.0295.13
Skip TuningShot=2, Time (s)=165, Mem. (M)=5282024.1295.1
CoCoOpShot=8, Time (s)=5358, Mem. (M)=28882024.1295.04
CS-AlignerBackbone=ViT-B/16, Number of Shots=16, Evaluation Protocol=fine-tuned2025.0295
TOGAShots=2, Venue=-2026.0395
CoCoOpShot=4, Time (s)=2684, Mem. (M)=28882024.1294.98
VaMPCross-dataset evaluation protocol=true, Zero-shot transfer=true2025.1194.96
MMA + AmPLevariant=standard2025.1294.9
MMRL + MetaTPTadaptation=MetaTPT2025.1294.9
CoCoOpShot=2, Time (s)=1346, Mem. (M)=28882024.1294.82
CLIP + MetaTPTadaptation=MetaTPT2025.1294.81
DPE*Setting=Inductive, Test-Time Adaptation (TTA)=true2025.0694.8
MMRLvariant=†2025.1294.77
MaPLe + AmPLevariant=standard2025.1294.77
MMRLCross-dataset evaluation protocol=true, Zero-shot transfer=true2025.1194.67
MMRLvariant=standard2025.1294.67
MMRLadaptation=none2025.1294.67
CAPTrole=Target2026.0394.62
UPLmode=unsupervised fine-tuning2026.0294.56
ProSRCShot=2, Time (s)=1947, Mem. (M)=33732024.1294.53
TACZero-shot=true, Source=ImageNet2026.0294.53
TACrole=Target2026.0394.53
LwEIBrole=Target2026.0394.51
CoPromptZero-shot=true, Source=ImageNet2026.0294.5
MaPLeShot=4, Time (s)=3541, Mem. (M)=30612024.1294.43
CoOpOpCross-dataset evaluation protocol=true, Zero-shot transfer=true2025.1194.43
CoCoOpvariant=standard2025.1294.43
CoCoOpZero-shot=true, Source=ImageNet2026.0294.43
CoOpShot=4, Time (s)=1036, Mem. (M)=45512024.1294.4
CoFT+mode=unsupervised fine-tuning2026.0294.4
MMRL + TPTadaptation=TPT2025.1294.4
CLIP-AdapterBackbone=ViT-B/16, Number of Shots=16, Evaluation Protocol=fine-tuned2025.0294.4
ZeroSetting=Inductive, Test-Time Adaptation (TTA)=false2025.0694.4
CoCoOpTraining shots=16-shot, Source dataset=ImageNet, Training epochs=32026.0394.4
CoOpShot=8, Time (s)=1988, Mem. (M)=45512024.1294.37
IPOvariant=standard2025.1294.34
MaPLe + MetaTPTadaptation=MetaTPT2025.1294.31
TAPZero-shot=true, Source=ImageNet2026.0294.3
TOGAShots=1, Venue=-2026.0394.3
AttriPromptvariant=standard2025.1294.23
MTA*Setting=Inductive, Test-Time Adaptation (TTA)=true2025.0694.2
TDA*Setting=Inductive, Test-Time Adaptation (TTA)=true2025.0694.2
CLIP + TPTadaptation=TPT2025.1294.16
CLIP-MoE + Det-FTTraining shots=16-shot, Source dataset=ImageNet, Training epochs=32026.0394.1
MaPLe + PromptAlignadaptation=PromptAlign2025.1294.01
MTABatch Size=1000, Backbone=ViT-B/16, Protocol=Batch test-time adaptation2025.0194
MaPLeShot=2, Time (s)=1785, Mem. (M)=30612024.1293.97
TCPCross-dataset evaluation protocol=true, Zero-shot transfer=true2025.1193.97
OursAnnotation Budget (B%)=1%, Backbone=CLIP ViT-B/16, Protocol=Full Framework (Joint Training)2026.0293.9
TIPPLE*Setting=Inductive, Test-Time Adaptation (TTA)=true2025.0693.9
Skip TuningShot=1, Time (s)=106, Mem. (M)=5282024.1293.87
CoCoOpShot=1, Time (s)=678, Mem. (M)=28882024.1293.83
MMACross-dataset evaluation protocol=true, Zero-shot transfer=true2025.1193.8
MMAvariant=standard2025.1293.8
MaPLevariant=†2025.1293.73
StatABatch Size=1000, Backbone=ViT-B/16, Keff=Very High (50-100), Protocol=Batch test-time adaptation2025.0193.7
CoOpCross-dataset evaluation protocol=true, Zero-shot transfer=true2025.1193.7
CoOpvariant=standard2025.1293.7
CoOpadaptation=none2025.1293.7
CoOpZero-shot=true, Source=ImageNet2026.0293.7
CoOprole=Target2026.0393.7
CoOpTraining shots=16-shot, Source dataset=ImageNet, Training epochs=32026.0393.7