List your company
ResearchFormerly WizWand

Medical Image Classification on NCT-CRC-HE-100K (test)

93.3AccuracyClean RAG (avg.)
-3.73221.45946.6571.841Aug 8, 2023Jan 22, 2024Jul 8, 2024Dec 23, 2024Jun 9, 2025Nov 24, 2025May 11, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Clean RAG (avg.)Model=GPT-4o, Retriever=Average2026.0593.3———
Clean RAG (avg.)Model=GPT-5, Retriever=Average2026.0592.02———
HistoDiffusionAugmentation Strategy=selective, Augmentation Ratio=200%, Classifier Backbone=ViT-B/162023.0891.991.691.999.2
HistoDiffusionAugmentation Strategy=selective, Augmentation Ratio=300%, Classifier Backbone=ViT-B/162023.089191.29198.8
Clean RAG (avg.)Model=Gemini-2.5, Retriever=Average2026.0590.45———
HistoDiffusionAugmentation Strategy=selective, Augmentation Ratio=100%, Classifier Backbone=ViT-B/162023.0890.390.290.399.1
StyleGAN2Augmentation Strategy=selective, Augmentation Ratio=300%, Classifier Backbone=ViT-B/162023.0889.889.689.898.7
LIARModel=GPT-4o, Retriever=CLIP2026.0589.67———
StyleGAN2Augmentation Strategy=selective, Augmentation Ratio=200%, Classifier Backbone=ViT-B/162023.0889.588.889.598.3
LIARModel=Gemini-2.5, Retriever=CLIP2026.0589.31———
HistoDiffusionAugmentation Strategy=selective, Augmentation Ratio=50%, Classifier Backbone=ViT-B/162023.0889.188.889.198.3
LIARModel=GPT-5, Retriever=CLIP2026.0588.46———
StyleGAN2Augmentation Strategy=selective, Augmentation Ratio=100%, Classifier Backbone=ViT-B/162023.0887.987.687.998.2
HistoDiffusionAugmentation Strategy=random, Augmentation Ratio=50%, Classifier Backbone=ViT-B/162023.088786.98798.2
StyleGAN2Augmentation Strategy=selective, Augmentation Ratio=50%, Classifier Backbone=ViT-B/162023.0886.886.186.797.8
StyleGAN2Augmentation Strategy=random, Augmentation Ratio=50%, Classifier Backbone=ViT-B/162023.088685.68698
BaselineData Split=5% real images, Classifier Backbone=ViT-B/162023.0885.58585.598.3
LIARModel=Gemini-2.5, Retriever=BGE-VL2026.0580.67———
M³AttModel=Gemini-2.5, Retriever=CLIP2026.0579.85———
LIARModel=Claude-4.5, Retriever=CLIP2026.0578.98———
LIARModel=GPT-4o, Retriever=BGE-VL2026.0578.83———
M³AttModel=GPT-4o, Retriever=CLIP2026.0578.41———
M³AttModel=GPT-5, Retriever=CLIP2026.0577.35———
LIARModel=GPT-5, Retriever=BGE-VL2026.0576.59———
Clean RAG (avg.)Model=Claude-4.5, Retriever=Average2026.0576.38———
LIARModel=LLaVA-Med, Retriever=CLIP2026.0575.63———
M³AttModel=Gemini-2.5, Retriever=BGE-VL2026.0573.69———
M³AttModel=GPT-4o, Retriever=BGE-VL2026.0570.62———
LIARModel=Claude-4.5, Retriever=BGE-VL2026.0570.18———
Clean RAG (avg.)Model=LLaVA-Med, Retriever=Average2026.0569.43———
M³AttModel=LLaVA-Med, Retriever=CLIP2026.0569.43———
M³AttModel=Claude-4.5, Retriever=CLIP2026.0569.28———
M³AttModel=GPT-5, Retriever=BGE-VL2026.0568.58———
M³AttModel=Claude-4.5, Retriever=BGE-VL2026.0563.02———
w/o RAGModel=Gemini-2.5, Retriever=None2026.0554.29———
LIARModel=LLaVA-Med, Retriever=BGE-VL2026.0553.22———
M³AttModel=LLaVA-Med, Retriever=BGE-VL2026.0550.16———
w/o RAGModel=GPT-4o, Retriever=None2026.0546.66———
w/o RAGModel=GPT-5, Retriever=None2026.0545.39———
LIARModel=LLaVA-Med, Retriever=Filtered2026.0525.63———
M³AttModel=LLaVA-Med, Retriever=Filtered2026.0523.81———
w/o RAGModel=LLaVA-Med, Retriever=None2026.0521.27———
w/o RAGModel=Claude-4.5, Retriever=None2026.0512.38———
LIARModel=Gemini-2.5, Retriever=Filtered2026.0511.39———
M³AttModel=Gemini-2.5, Retriever=Filtered2026.059.84———
LIARModel=GPT-4o, Retriever=Filtered2026.050———
M³AttModel=GPT-4o, Retriever=Filtered2026.050———
LIARModel=GPT-5, Retriever=Filtered2026.050———
M³AttModel=GPT-5, Retriever=Filtered2026.050———
LIARModel=Claude-4.5, Retriever=Filtered2026.050———
M³AttModel=Claude-4.5, Retriever=Filtered2026.050———