Image Classification on PCam (test)
90.76AccuracyDINO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DINOBackbone=ViT-S, Protocol=Fine-tune, Patch Size=82022.12 | 90.76 | — | |
| DINOBackbone=ViT-S, Protocol=Fine-tune, Patch Size=162022.12 | 90.4 | — | |
| DINOBackbone=ViT-S, Protocol=Linear, Patch Size=82022.12 | 90.12 | — | |
| DINOBackbone=ViT-S, Protocol=Linear, Patch Size=162022.12 | 88.78 | — | |
| SupervisedBackbone=ViT-S, Protocol=Fine-tune, Patch Size=162022.12 | 88.3 | — | |
| BTBackbone=ResNet-50, Protocol=Linear2022.12 | 88.15 | — | |
| MoCo v2Backbone=ResNet-50, Protocol=Fine-tune2022.12 | 87.62 | — | |
| SwAVBackbone=ResNet-50, Protocol=Fine-tune2022.12 | 87.6 | — | |
| BTBackbone=ResNet-50, Protocol=Fine-tune2022.12 | 86.92 | — | |
| MoCo v2Backbone=ResNet-50, Protocol=Linear2022.12 | 86.78 | — | |
| SwAVBackbone=ResNet-50, Protocol=Linear2022.12 | 85.28 | — | |
| Supervised CLIPTraining Protocol=Supervised2021.06 | 82.6 | — | |
| SupervisedBackbone=ViT-S, Protocol=Linear, Patch Size=162022.12 | 80.96 | — | |
| CLIP + EB βTraining Protocol=Zero-shot SSL compressor2021.06 | 80.9 | 64 | |
| SupervisedBackbone=ResNet-50, Protocol=Linear2022.12 | 80.79 | — | |
| SupervisedBackbone=ResNet-50, Protocol=Fine-tune2022.12 | 80.63 | — | |
| RandomBackbone=ResNet-50, Protocol=Linear2022.12 | 76.52 | — | |
| RandomBackbone=ResNet-50, Protocol=Fine-tune2022.12 | 75.71 | — | |
| RandomBackbone=ViT-S, Protocol=Fine-tune, Patch Size=162022.12 | 75.42 | — | |
| LIARModel=Gemini-2.5, Retriever=CLIP2026.05 | 75.36 | — | |
| Clean RAG (avg.)Model=Gemini-2.5, Retriever=Average2026.05 | 74.58 | — | |
| RandomBackbone=ViT-S, Protocol=Linear, Patch Size=162022.12 | 74.43 | — | |
| LIARModel=Gemini-2.5, Retriever=BGE-VL2026.05 | 73.58 | — | |
| LIARModel=GPT-4o, Retriever=CLIP2026.05 | 71.5 | — | |
| CONCHModel=CONCH, Adaptation Strategy=Original Zero-shot2026.06 | 69.95 | — | |
| Clean RAG (avg.)Model=GPT-4o, Retriever=Average2026.05 | 66.62 | — | |
| MUSKModel=MUSK, Adaptation Strategy=Original Zero-shot2026.06 | 66 | — | |
| M³AttModel=Gemini-2.5, Retriever=BGE-VL2026.05 | 65.76 | — | |
| KEEPModel=KEEP, Adaptation Strategy=Original Zero-shot2026.06 | 65.68 | — | |
| PathGen-CLIPModel=PathGen-CLIP, Adaptation Strategy=Original Zero-shot2026.06 | 64.89 | — | |
| QuiltNet-B/16Model=QuiltNet-B/16, Adaptation Strategy=Original Zero-shot2026.06 | 63.92 | — | |
| BiomedCLIPModel=BiomedCLIP, Adaptation Strategy=single-VLM GMM2026.06 | 63.7 | — | |
| CONCHModel=CONCH, Adaptation Strategy=single-VLM GMM2026.06 | 62.35 | — | |
| BiomedCLIPModel=BiomedCLIP, Adaptation Strategy=Original Zero-shot2026.06 | 61.14 | — | |
| M³AttModel=Gemini-2.5, Retriever=CLIP2026.05 | 59.44 | — | |
| KEEPModel=KEEP, Adaptation Strategy=single-VLM GMM2026.06 | 59.06 | — | |
| KEEPModel=KEEP, Adaptation Strategy=multi-VLM GMM2026.06 | 59.05 | — | |
| LIARModel=Claude-4.5, Retriever=CLIP2026.05 | 59 | — | |
| QuiltNet-B/32Model=QuiltNet-B/32, Adaptation Strategy=Original Zero-shot2026.06 | 58.31 | — | |
| CLIP-B/32Model=CLIP-B/32, Adaptation Strategy=single-VLM GMM2026.06 | 57.74 | — | |
| MUSKModel=MUSK, Adaptation Strategy=multi-VLM GMM2026.06 | 57.46 | — | |
| MUSKModel=MUSK, Adaptation Strategy=single-VLM GMM2026.06 | 56.88 | — | |
| PathGen-CLIPModel=PathGen-CLIP, Adaptation Strategy=single-VLM GMM2026.06 | 56.43 | — | |
| OSTBModel=OSTB, Adaptation Strategy=Adapted Ensemble2026.06 | 56.31 | — | |
| PathGen-CLIPModel=PathGen-CLIP, Adaptation Strategy=multi-VLM GMM2026.06 | 55.98 | — | |
| CONCHModel=CONCH, Adaptation Strategy=multi-VLM GMM2026.06 | 55.89 | — | |
| UniMed-CLIPModel=UniMed-CLIP, Adaptation Strategy=multi-VLM GMM2026.06 | 55.81 | — | |
| PLIPModel=PLIP, Adaptation Strategy=Original Zero-shot2026.06 | 55.19 | — | |
| CLIP-B/32Model=CLIP-B/32, Adaptation Strategy=Original Zero-shot2026.06 | 54.77 | — | |
| LIARModel=GPT-4o, Retriever=BGE-VL2026.05 | 54.54 | — | |
| QuiltNet-B/32Model=QuiltNet-B/32, Adaptation Strategy=multi-VLM GMM2026.06 | 54.2 | — | |
| Clean RAG (avg.)Model=Claude-4.5, Retriever=Average2026.05 | 53.92 | — | |
| PLIPModel=PLIP, Adaptation Strategy=multi-VLM GMM2026.06 | 53.12 | — | |
| QuiltNet-B/16Model=QuiltNet-B/16, Adaptation Strategy=single-VLM GMM2026.06 | 52.6 | — | |
| BiomedCLIPModel=BiomedCLIP, Adaptation Strategy=multi-VLM GMM2026.06 | 52.45 | — | |
| QuiltNet-B/16Model=QuiltNet-B/16, Adaptation Strategy=multi-VLM GMM2026.06 | 51.57 | — | |
| PMC-CLIPModel=PMC-CLIP, Adaptation Strategy=Original Zero-shot2026.06 | 50.9 | — | |
| CLIP-B/16Model=CLIP-B/16, Adaptation Strategy=single-VLM GMM2026.06 | 50.79 | — | |
| UniMed-CLIPModel=UniMed-CLIP, Adaptation Strategy=Original Zero-shot2026.06 | 50.77 | — | |
| CLIP-B/32Model=CLIP-B/32, Adaptation Strategy=multi-VLM GMM2026.06 | 50.67 | — | |
| PLIPModel=PLIP, Adaptation Strategy=single-VLM GMM2026.06 | 50.26 | — | |
| M³AttModel=GPT-4o, Retriever=CLIP2026.05 | 50.02 | — | |
| QuiltNet-B/32Model=QuiltNet-B/32, Adaptation Strategy=single-VLM GMM2026.06 | 49.68 | — | |
| CLIP-B/16Model=CLIP-B/16, Adaptation Strategy=multi-VLM GMM2026.06 | 49.61 | — | |
| PMC-CLIPModel=PMC-CLIP, Adaptation Strategy=multi-VLM GMM2026.06 | 49.6 | — | |
| M³AttModel=Claude-4.5, Retriever=CLIP2026.05 | 49.14 | — | |
| CLIP-B/16Model=CLIP-B/16, Adaptation Strategy=Original Zero-shot2026.06 | 47.13 | — | |
| M³AttModel=GPT-4o, Retriever=BGE-VL2026.05 | 45.5 | — | |
| UniMed-CLIPModel=UniMed-CLIP, Adaptation Strategy=single-VLM GMM2026.06 | 45.39 | — | |
| LIARModel=Claude-4.5, Retriever=BGE-VL2026.05 | 45.22 | — | |
| w/o RAGModel=Gemini-2.5, Retriever=None2026.05 | 43.44 | — | |
| M³AttModel=Claude-4.5, Retriever=BGE-VL2026.05 | 41.12 | — | |
| LIARModel=GPT-5, Retriever=BGE-VL2026.05 | 40.7 | — | |
| PMC-CLIPModel=PMC-CLIP, Adaptation Strategy=single-VLM GMM2026.06 | 37.59 | — | |
| Clean RAG (avg.)Model=GPT-5, Retriever=Average2026.05 | 37.24 | — | |
| M³AttModel=GPT-5, Retriever=BGE-VL2026.05 | 36.52 | — | |
| LIARModel=GPT-5, Retriever=CLIP2026.05 | 33.86 | — | |
| M³AttModel=GPT-5, Retriever=CLIP2026.05 | 27.5 | — | |
| LIARModel=LLaVA-Med, Retriever=Filtered2026.05 | 22 | — | |
| w/o RAGModel=Claude-4.5, Retriever=None2026.05 | 21.22 | — | |
| M³AttModel=LLaVA-Med, Retriever=Filtered2026.05 | 15.16 | — | |
| w/o RAGModel=GPT-5, Retriever=None2026.05 | 13.14 | — | |
| LIARModel=Claude-4.5, Retriever=Filtered2026.05 | 12.3 | — | |
| M³AttModel=LLaVA-Med, Retriever=CLIP2026.05 | 10.48 | — | |
| LIARModel=LLaVA-Med, Retriever=CLIP2026.05 | 9.78 | — | |
| LIARModel=GPT-5, Retriever=Filtered2026.05 | 7.18 | — | |
| M³AttModel=Claude-4.5, Retriever=Filtered2026.05 | 7.08 | — | |
| Clean RAG (avg.)Model=LLaVA-Med, Retriever=Average2026.05 | 3.04 | — | |
| M³AttModel=LLaVA-Med, Retriever=BGE-VL2026.05 | 2.84 | — | |
| LIARModel=LLaVA-Med, Retriever=BGE-VL2026.05 | 2.5 | — | |
| M³AttModel=GPT-5, Retriever=Filtered2026.05 | 1.02 | — | |
| w/o RAGModel=GPT-4o, Retriever=None2026.05 | 0 | — | |
| LIARModel=GPT-4o, Retriever=Filtered2026.05 | 0 | — | |
| M³AttModel=GPT-4o, Retriever=Filtered2026.05 | 0 | — | |
| LIARModel=Gemini-2.5, Retriever=Filtered2026.05 | 0 | — | |
| M³AttModel=Gemini-2.5, Retriever=Filtered2026.05 | 0 | — | |
| w/o RAGModel=LLaVA-Med, Retriever=None2026.05 | 0 | — |