Recent rapid progress in the field of computational pathology has been enabled by foundation models. These models are beginning to move beyond encoding image patches towards whole-slide understanding but their clinical utility remains limited. In this work, we present PRISM2, a multimodal slide-level foundation model trained on data from 700,000 diagnostic specimen-report pairs, the largest vision (2.3 million whole slide images) and language (14M question-answer pairs) histopathology dataset to date. By learning through clinical-dialogue supervision, PRISM2 aligns histomorphologic features with the language of diagnostic reasoning, producing slide-level representations that support both direct diagnostic question-answering and transferable embeddings for downstream tasks. Without additional training, PRISM2 matches or exceeds the cancer-detection performance of clinical-grade products. This is observed without loss of generality on other tasks, where PRISM2 achieves top performance. Finally, using survival prediction as the example, we show that task-specific finetuning with a large dataset can outperform task-specific models, further improving performance. These results demonstrate how language-supervised pretraining provides a scalable, clinically grounded signal for learning generalizable pathology representations, bridging human diagnostic reasoning and foundation-model performance.
@article{arxiv.2506.13063,
title = {PRISM2: Unlocking Multi-Modal General Pathology AI with Clinical Dialogue},
author = {Eugene Vorontsov and George Shaikovski and Adam Casson and Julian Viret and Eric Zimmermann and Neil Tenenholtz and Yi Kan Wang and Jan H. Bernhard and Ran A. Godrich and Juan A. Retamero and Jinru Shia and Mithat Gonen and Martin R. Weiser and David S. Klimstra and Razik Yousfi and Nicolo Fusi and Thomas J. Fuchs and Kristen Severson and Siqi Liu},
journal= {arXiv preprint arXiv:2506.13063},
year = {2025}
}