List your company
ResearchFormerly WizWand

Image Classification on MHIST (test)

94.27AccuracyRMSGD
-3.770821.682147.13572.5879Mar 31, 2022Dec 6, 2022Aug 13, 2023Apr 20, 2024Dec 26, 2024Sep 2, 2025May 11, 2026
Updated 5mo ago

Evaluation results

MethodLinks
RMSGDNetwork=ResNet182022.0394.27
RMSGDNetwork=ResNet342022.0394.19
AdamPNetwork=ResNet182022.0394.04
AdamPNetwork=ResNet342022.0393.95
RMSGDNetwork=MobileNetV22022.0393.83
AdamPNetwork=MobileNetV22022.0393.78
SGDPNetwork=ResNet342022.0393.65
SGDPNetwork=ResNet182022.0393.48
SGDNetwork=ResNet342022.0393.38
SAMNetwork=ResNet182022.0393.28
SAMNetwork=ResNet342022.0393.24
SGDNetwork=ResNet182022.0393.22
AdamNetwork=MobileNetV22022.0392.89
AdamNetwork=ResNet342022.0392.8
AdamNetwork=ResNet182022.0392.75
SGDPNetwork=MobileNetV22022.0391.42
SAMNetwork=MobileNetV22022.0389.43
SGDNetwork=MobileNetV22022.0388.81
MoCo v2Backbone=ResNet-50, Protocol=Fine-tune2022.1285.88
SupervisedBackbone=ViT-S, Protocol=Fine-tune, Patch Size=162022.1281.68
BTBackbone=ResNet-50, Protocol=Fine-tune2022.1281.27
DINOBackbone=ViT-S, Protocol=Fine-tune, Patch Size=162022.1279.43
SupervisedBackbone=ResNet-50, Protocol=Fine-tune2022.1278.92
BTBackbone=ResNet-50, Protocol=Linear2022.1278.81
SupervisedBackbone=ViT-S, Protocol=Linear, Patch Size=162022.1278.51
DINOBackbone=ViT-S, Protocol=Fine-tune, Patch Size=82022.1278.4
SwAVBackbone=ResNet-50, Protocol=Fine-tune2022.1277.99
DINOBackbone=ViT-S, Protocol=Linear, Patch Size=82022.1277.89
MoCo v2Backbone=ResNet-50, Protocol=Linear2022.1277.07
SupervisedBackbone=ResNet-50, Protocol=Linear2022.1276.25
DINOBackbone=ViT-S, Protocol=Linear, Patch Size=162022.1276.15
RandomBackbone=ResNet-50, Protocol=Fine-tune2022.1275.54
SwAVBackbone=ResNet-50, Protocol=Linear2022.1271.14
Clean RAG (avg.)Model=LLaVA-Med, Retriever=Average2026.0570.77
Clean RAG (avg.)Model=GPT-4o, Retriever=Average2026.0570.2
Clean RAG (avg.)Model=GPT-5, Retriever=Average2026.0569.21
Clean RAG (avg.)Model=Claude-4.5, Retriever=Average2026.0567.57
LIARModel=LLaVA-Med, Retriever=BGE-VL2026.0566.19
RandomBackbone=ViT-S, Protocol=Linear, Patch Size=162022.1263.46
RandomBackbone=ResNet-50, Protocol=Linear2022.1263.15
RandomBackbone=ViT-S, Protocol=Fine-tune, Patch Size=162022.1262.13
LIARModel=GPT-5, Retriever=BGE-VL2026.0561.61
LIARModel=GPT-5, Retriever=CLIP2026.0560.71
LIARModel=Claude-4.5, Retriever=BGE-VL2026.0558.97
ACE-LoRASetting=Zero-shot2026.0357.27
LIARModel=LLaVA-Med, Retriever=CLIP2026.0557.14
QuiltNetSetting=Zero-shot2026.0356.82
PLIPSetting=Zero-shot2026.0356.63
LIARModel=GPT-4o, Retriever=BGE-VL2026.0555.99
M³AttModel=LLaVA-Med, Retriever=BGE-VL2026.0554.76
LIARModel=GPT-4o, Retriever=CLIP2026.0554.27
M³AttModel=LLaVA-Med, Retriever=CLIP2026.0549.4
LIARModel=Claude-4.5, Retriever=CLIP2026.0549.11
M³AttModel=GPT-5, Retriever=CLIP2026.0548.87
M³AttModel=GPT-5, Retriever=BGE-VL2026.0548.64
LIARModel=LLaVA-Med, Retriever=Filtered2026.0544.11
BMC-CLIPSetting=Zero-shot2026.0341.78
M³AttModel=Claude-4.5, Retriever=BGE-VL2026.0540.84
M³AttModel=GPT-4o, Retriever=CLIP2026.0539.64
M³AttModel=GPT-4o, Retriever=BGE-VL2026.0539
M³AttModel=LLaVA-Med, Retriever=Filtered2026.0538.57
w/o RAGModel=LLaVA-Med, Retriever=None2026.0537.14
M³AttModel=Claude-4.5, Retriever=CLIP2026.0536.24
LIARModel=GPT-5, Retriever=Filtered2026.0534.67
BiomedCLIPSetting=Zero-shot2026.0334.04
w/o RAGModel=GPT-5, Retriever=None2026.0531.43
M³AttModel=GPT-5, Retriever=Filtered2026.0530
LIARModel=Gemini-2.5, Retriever=CLIP2026.0527.64
M³AttModel=Gemini-2.5, Retriever=CLIP2026.0521.93
LIARModel=Gemini-2.5, Retriever=Filtered2026.0517.06
LIARModel=Claude-4.5, Retriever=Filtered2026.0516.96
LIARModel=GPT-4o, Retriever=Filtered2026.0516.63
M³AttModel=GPT-4o, Retriever=Filtered2026.0514.29
M³AttModel=Claude-4.5, Retriever=Filtered2026.0514.29
M³AttModel=Gemini-2.5, Retriever=Filtered2026.0513.57
w/o RAGModel=GPT-4o, Retriever=None2026.0512.14
w/o RAGModel=Gemini-2.5, Retriever=None2026.0512.14
w/o RAGModel=Claude-4.5, Retriever=None2026.0512.14
Clean RAG (avg.)Model=Gemini-2.5, Retriever=Average2026.053.14
LIARModel=Gemini-2.5, Retriever=BGE-VL2026.050
M³AttModel=Gemini-2.5, Retriever=BGE-VL2026.050