[R] Google DeepMind Diagnostic LLM Exceeds Human Doctor Top-10 Accuracy (59% vs 34%)

Researchers from Google and DeepMind have developed and evaluated an LLM fine-tuned specifically for clinical diagnostic reasoning. In a new study, they rigorously tested the LLM's aptitude for generating differential diagnoses and aiding physicians.

They assessed the LLM on 302 real-world case reports from the New England Journal of Medicine. These case reports are known to be highly complex diagnostic challenges.

The LLM produced differential diagnosis lists that included the final confirmed diagnosis in the top 10 possibilities in 177 …

accuracy case clinical deepmind diagnostic differential doctor england google google deepmind human llm machinelearning physicians reasoning reports researchers study world