Tempus Cancer Model Boosts Survival Prediction AUC with 1.67M Patient Dataset
neuroecology · x · 2026-08-27
Tempus introduced the oFM (Oncology Foundation Model), a multimodal foundation model trained on a real-world cohort of 1.67 million cancer patients. It fuses clinical trajectories with DNA, RNA, and H&E pathology images into a unified patient-state embedding.
Key Results:
- In prognostic benchmarks, the oFM improved the Area Under the Curve (AUC) for overall survival prediction from 0.563 to 0.774, significantly outperforming expert-curated clinical and molecular feature baselines.
- Across 11 comparative-treatment cohorts, oFM embeddings improved treatment-benefit ranking in 9 cohorts, achieving a three-fold higher pooled and scale-normalized treatment-benefit AUTOC than baseline features.
Technical Details:
- The model encodes daily clinical and molecular episodes and integrates them over time alongside pathology images.
- The study utilized over one million patients for training with strict patient-level partitioning for validation and testing.
More from Research
- Interpretability paper incoming: in-depth look at testing explanations on modern LLMs — burny_tech · 2026-10-16
- Ex-OpenAI interpretability lead: neurons in trillion-param models are mostly legible now — CatAstro_Piyush · 2026-10-16
- Better-learning models tend to be more interpretable, argues Nick Cammarata — burny_tech · 2026-10-16
- MIT let Claude Code loose on a virtual island: it played for 30 hours with no task — CatAstro_Piyush · 2026-10-16
- Cornell and Amazon revisit the 16x pretraining compute gap: diffusion LMs need only 3-8x — LucaAmb · 2026-10-16
- AI trained on the fraction of human intelligence that fit the page — r0ck3t23 · 2026-10-16