List your company
ResearchFormerly WizWand

Image Classification on PCAM

95.9Top-1 AccCPath-CLIP
48.26860.6347385.366Nov 19, 2021Aug 19, 2022May 19, 2023Feb 16, 2024Nov 15, 2024Aug 15, 2025May 16, 2026
Updated 5mo ago

Evaluation results

MethodLinks
CPath-CLIPzero-shot=true2024.1295.9
FLYPEvaluation Protocol=Fine-tuning2022.1290.3
FTEvaluation Protocol=Fine-tuning2022.1289.1
LP-FTEvaluation Protocol=Linear Probing then Fine-tuning2022.1289
PathGen-CLIP-Lzero-shot=true2024.1288.2
WUDI-Merging w/ E-PMQMerging Strategy=WUDI-Merging, Quantization Method=E-PMQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0586.57
WUDI-Merging + E-PMQMerging Strategy=WUDI-Merging, Quantization Method=E-PMQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0585.9
WUDI-MergingMerging Strategy=WUDI-Merging, Quantization Method=None, Model Backbone=CLIP-ViT-B/32, Bit-width=Full2026.0585.5
FLAVALinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.0185.3
WUDI-Merging w/ GPTQMerging Strategy=WUDI-Merging, Quantization Method=GPTQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0585.06
nCLIPLinear probing=true, Backbone=ViT-B/16, Pre-train=IT35M2022.1084.2
BiomedCLIPzero-shot=true2024.1284
DaVinciLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.0183.6
xCLIPLinear probing=true, Backbone=ViT-B/16, Pre-train=IT35M2022.1083.5
CLIPLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.0183.5
CLIPLinear probing=true, Backbone=ViT-B/16, Pre-train=IT35M2022.1083.4
LPEvaluation Protocol=Linear Probing2022.1282.6
WUDI-Merging w/ RTNMerging Strategy=WUDI-Merging, Quantization Method=RTN, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0582.48
X-FM_baseLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.0182.4
WUDI-MergingMerging Strategy=WUDI-Merging, Quantization Method=Full-precision, Bit-width=Full, Backbone=CLIP-ViT-B/322026.0582
WUDI-Merging w/ AWQMerging Strategy=WUDI-Merging, Quantization Method=AWQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0580.97
ResNet-18Backbone=ResNet-182026.0380.71
RoPEPositional Encoding=RoPE2026.0379.82
Learnable PEPositional Encoding=Learnable2026.0379.74
No PEPositional Encoding=None2026.0379.65
WUDI-Merging + GPTQMerging Strategy=WUDI-Merging, Quantization Method=GPTQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0577.44
TIES-MergingMerging Strategy=TIES-Merging, Quantization Method=None, Model Backbone=CLIP-ViT-B/32, Bit-width=Full2026.0577.1
Simple Averaging w/ E-PMQMerging Strategy=Simple Averaging, Quantization Method=E-PMQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0575.44
TIES-Merging w/ E-PMQMerging Strategy=TIES-Merging, Quantization Method=E-PMQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0574.39
Task Arithmetic w/ E-PMQMerging Strategy=Task Arithmetic, Quantization Method=E-PMQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0573.64
Simple Averaging + E-PMQMerging Strategy=Simple Averaging, Quantization Method=E-PMQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0573.57
TIES-Merging + E-PMQMerging Strategy=TIES-Merging, Quantization Method=E-PMQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0573.08
PathCLIPzero-shot=true2024.1272.5
WUDI-Merging + RTNMerging Strategy=WUDI-Merging, Quantization Method=RTN, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0572.16
DFN5B-CLIP-H/14+Zero-shot=true2024.0269.6
WUDI-Merging + AWQMerging Strategy=WUDI-Merging, Quantization Method=AWQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0567.75
Task Arithmetic + E-PMQMerging Strategy=Task Arithmetic, Quantization Method=E-PMQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0566.98
TIES-MergingMerging Strategy=TIES-Merging, Quantization Method=Full-precision, Bit-width=Full, Backbone=CLIP-ViT-B/322026.0566.9
DFN5B-CLIP-H/14Zero-shot=true2024.0265.2
EVA-CLIP-18BZero-shot=true2024.0265.2
Simple AveragingMerging Strategy=Simple Averaging, Quantization Method=None, Model Backbone=CLIP-ViT-B/32, Bit-width=Full2026.0565.2
EVA-02-CLIP-E/14+Zero-shot=true2024.0264.4
Simple Averaging w/ GPTQMerging Strategy=Simple Averaging, Quantization Method=GPTQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0564.39
Task ArithmeticMerging Strategy=Task Arithmetic, Quantization Method=None, Model Backbone=CLIP-ViT-B/32, Bit-width=Full2026.0564.1
Simple AveragingMerging Strategy=Simple Averaging, Quantization Method=Full-precision, Bit-width=Full, Backbone=CLIP-ViT-B/322026.0563.9
OpenCLIP-G/14Zero-shot=true2024.0263.6
CLIPBackbone=ViT-L/14-3362021.1163
CLIPPre-training Dataset=DataComp-1B, Backbone=ViT-B/16, Training Epochs=10 epochs, Zero-shot protocol=EVA-CLIP2026.0362.8
EVA-01-CLIP-g/14+Zero-shot=true2024.0262.4
TinyCLIPImage Encoder=ViT-39M/16, Zero-shot=true2026.0260.5
CLIPImage Encoder=ViT-B/16, Zero-shot=true2026.0260.2
TIES-Merging w/ GPTQMerging Strategy=TIES-Merging, Quantization Method=GPTQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0559.66
BASIC-LBackbone=BASIC-L2021.1159.6
ITO sub2Pre-train dataset=CC3M, Backbone=ViT-B/16, Training Epochs=30, Evaluation Protocol=Zero-shot2026.0359.6
CLIP (reported)Pre-training Data=WIT-400M, Backbone=ViT-L/14, Evaluation Protocol=Zero-shot classification2024.0758.8
QuiltNetzero-shot=true2024.1258.7
BASIC-MBackbone=BASIC-M2021.1158.3
Task Arithmetic w/ GPTQMerging Strategy=Task Arithmetic, Quantization Method=GPTQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0558.14
EVA-CLIP-8BZero-shot=true2024.0258.1
Task Arithmetic w/ AWQMerging Strategy=Task Arithmetic, Quantization Method=AWQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0557.72
ITOPre-training Dataset=DataComp-1B, Backbone=ViT-B/16, Training Epochs=10 epochs, Zero-shot protocol=EVA-CLIP2026.0357.7
Simple Averaging + GPTQMerging Strategy=Simple Averaging, Quantization Method=GPTQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0557.65
CLIPBackbone=ResNet-502021.1157.6
TIES-Merging + GPTQMerging Strategy=TIES-Merging, Quantization Method=GPTQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0557.45
CLIP-MapsmallImage Encoder=ViT-8M/16, Zero-shot=true2026.0257.4
FLAIRPre-train dataset=CC3M, Backbone=ViT-B/16, Training Epochs=30, Evaluation Protocol=Zero-shot2026.0357.2
ITO sub2Pre-training Dataset=DataComp-1B, Backbone=ViT-B/16, Training Epochs=10 epochs, Zero-shot protocol=EVA-CLIP2026.0356.5
ZeroshotEvaluation Protocol=Zero-shot2022.1256.49
CLIP-PGSImage Encoder=ViT-S/16, p=0.5, Protocol=Zero-shot2025.0356.4
OpenCLIPImage Encoder=ViT-B/16, Zero-shot=true2026.0256.4
SLIPPre-train dataset=CC3M, Backbone=ViT-B/16, Training Epochs=30, Evaluation Protocol=Zero-shot2026.0355.9
CLIPPre-train dataset=CC3M, Backbone=ViT-B/16, Training Epochs=30, Evaluation Protocol=Zero-shot2026.0355.6
OpenAI-CLIP-Lzero-shot=true2024.1255.5
BASIC-SBackbone=BASIC-S2021.1154.3
Task Arithmetic w/ RTNMerging Strategy=Task Arithmetic, Quantization Method=RTN, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0554.2
MLCDPre-training Data=LAION-400M, Backbone=ViT-L/14, Evaluation Protocol=Zero-shot classification2024.0753.7
Simple Averaging + AWQMerging Strategy=Simple Averaging, Quantization Method=AWQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0553.55
Simple Averaging w/ AWQMerging Strategy=Simple Averaging, Quantization Method=AWQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0553.3
Task ArithmeticMerging Strategy=Task Arithmetic, Quantization Method=Full-precision, Bit-width=Full, Backbone=CLIP-ViT-B/322026.0553.2
†TinyCLIPImage Encoder=ViT-8M/16, Zero-shot=true2026.0252.8
Task Arithmetic + AWQMerging Strategy=Task Arithmetic, Quantization Method=AWQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0552.62
TIES-Merging w/ RTNMerging Strategy=TIES-Merging, Quantization Method=RTN, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0552.49
Simple Averaging w/ RTNMerging Strategy=Simple Averaging, Quantization Method=RTN, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0552.38
TIES-Merging + AWQMerging Strategy=TIES-Merging, Quantization Method=AWQ, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0552.19
PLIPzero-shot=true2024.1251.8
CLIP-MapbaseImage Encoder=ViT-39M/16, Zero-shot=true2026.0251.7
CLIPPre-training Data=WIT-400M, Backbone=ViT-L/14, Evaluation Protocol=Zero-shot classification2024.0751.6
TIES-Merging w/ AWQMerging Strategy=TIES-Merging, Quantization Method=AWQ, Model Backbone=CLIP-ViT-B/32, Bit-width=4-bit2026.0551.56
nCLIPBackbone=ViT-B/16, Pre-trained on=IT35M, Mode=Zero-shot2022.1051.2
xCLIPBackbone=ViT-B/16, Pre-trained on=IT35M, Mode=Zero-shot2022.1051.1
OpenCLIPPre-training Data=LAION-400M, Backbone=ViT-L/14, Evaluation Protocol=Zero-shot classification2024.0751.1
Simple Averaging + RTNMerging Strategy=Simple Averaging, Quantization Method=RTN, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0551.07
CLIPZero-shot=true, Backbone=ViT-B/16, Pre-training Dataset=Laion100M, Epochs=302026.0351
CLIP-PGSImage Encoder=ViT-B/16, p=0.3, Protocol=Zero-shot2025.0350.8
TIES-Merging + RTNMerging Strategy=TIES-Merging, Quantization Method=RTN, Bit-width=4-bit, Backbone=CLIP-ViT-B/322026.0550.61
FLIPPre-training Data=LAION-400M, Backbone=ViT-L/14, Evaluation Protocol=Zero-shot classification2024.0750.3
SLIPZero-shot=true, Backbone=ViT-B/16, Pre-training Dataset=Laion100M, Epochs=302026.0350.3
ITOZero-shot=true, Backbone=ViT-B/16, Pre-training Dataset=Laion100M, Epochs=302026.0350.3
CLIP-PGSImage Encoder=ViT-S/16, p=0.3, Protocol=Zero-shot2025.0350.2
A-CLIPImage Encoder=ViT-B/16, Protocol=Zero-shot2025.0350.1