Slide-level Pathology Diagnosis on SlideBench BCNB
59.42Overall AccuracyPathNavigate
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| PathNavigateCategory=Training-Free Pathology Agents2026.05 | 59.42 | 88.28 | 76.09 | 69 | 48.4 | 52.58 | 32.33 | |
| PathAgentCategory=Training-Free Pathology Agents2026.05 | 55.72 | 87.52 | 63.69 | 63.69 | 44.52 | 55.95 | 30.21 | |
| SlideChatCategory=Supervised Pathology Models2026.05 | 54.07 | 89.9 | 72.98 | 72.98 | 47.53 | 23.11 | 20.58 | |
| WSI-LLaVACategory=Supervised Pathology Models2026.05 | 52.68 | 85.32 | 60.81 | 60.81 | 42.78 | 46.28 | 29.2 | |
| MedGemma 1.5-4BCategory=Supervised Pathology Models2026.05 | 50.38 | 79.68 | 52.36 | 60.35 | 47.85 | 42.33 | 21.27 | |
| GPT-4.5Category=Pretrained MLLM Models2026.05 | 45.85 | 60.42 | 50.42 | 49.86 | 44.51 | 42.36 | 28.43 | |
| Gemini-3.1-Pro-previewCategory=Pretrained MLLM Models2026.05 | 44.91 | 58.46 | 50.18 | 49.62 | 43.58 | 40.85 | 27.62 | |
| TITANCategory=Supervised Pathology Models, Zero-shot=true, Feature extractor=CONCH v1.5 feats2026.05 | 44.39 | 85.16 | 42.44 | 41.68 | 36.86 | 45.35 | 22.49 | |
| GPT-4oCategory=Pretrained MLLM Models2026.05 | 43.78 | 56.84 | 49.96 | 49.42 | 42.13 | 38.46 | 26.84 | |
| Qwen3.5-9BCategory=Pretrained MLLM Models2026.05 | 41.99 | 53.84 | 36.42 | 30.18 | 47.53 | 59.15 | 21.43 | |
| Qwen3.5-4BCategory=Pretrained MLLM Models2026.05 | 39.48 | 49.43 | 33.36 | 27.32 | 48.02 | 52.7 | 19.19 | |
| Patho-R1 aloneCategory=Supervised Pathology Models, Agent integration=no agent, Context=PathNavigate’s perceptor2026.05 | 37.26 | 42.16 | 54.63 | 50.38 | 25.62 | 35.53 | 26.65 | |
| MedDrCategory=Supervised Pathology Models2026.05 | 35.48 | 59.6 | 43.8 | 39.2 | 27.4 | 29.4 | 20.8 | |
| Quilt-LLaVACategory=Supervised Pathology Models2026.05 | 34.11 | 33.46 | 49.43 | 44.05 | 25.85 | 34.99 | 25.24 |