Virchow2 has been ranked #1 on the Eva leaderboard, a leading evaluation framework for oncology foundation models by kaiko.ai. Eva rigorously assessed 13 foundation models on 6 WSI-classification tasks using a fixed protocol that prioritizes fairness and comparability over state-of-the-art performance. Virchow2 achieved state-of-the-art performance across classification and segmentation tasks for various tissue types, including breast, colorectal, kidney, lung, and prostate. What sets Virchow2 apart? ☑️ Trained on 3M whole slide images of H&E and IHC samples ☑️ Learning across 5x, 10x, 20x, and 40x magnifications ☑️ Available as a commercial or an open-weights model via Hugging Face and Azure AI Studio Explore Virchow2 and learn why it’s setting a new standard in the field: 📈 Leaderboard: https://lnkd.in/ghUY6zyR 💻 Hugging Face: https://lnkd.in/gTG_j3bw ☁️ Azure AI Foundry: aka.ms/Virchow2
Thanks for sharing
Research Engineer @ Paige AI (vision-language models)
2wAlso huge thanks to kaiko.ai for maintaining Eva and the leaderboard - no matter who’s at the top, having centralised unbiased evaluations are very important for the community!