What question did this study set out to answer?

This research aims to enhance medical question answering by improving large language models, focusing on real-world applicability.

January 8, 2025Nature MedicineOpen Access

Toward expert-level medical question answering with large language models

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial shows Med-PaLM 2 outperforms physician responses in medical question answering, suggesting enhanced clinical support.

This research aims to enhance medical question answering by improving large language models, focusing on real-world applicability.
Developed Med-PaLM 2 with base model improvements and domain-specific fine-tuning.
Employed ensemble refinement and chain of retrieval for enhanced reasoning.
Conducted detailed human evaluations to compare Med-PaLM 2 with physician responses.
Med-PaLM 2 scored up to 86.5% on MedQA, surpassing Med-PaLM by over 19%.
Physicians preferred Med-PaLM 2 answers on eight of nine clinical axes.
Specialists favored Med-PaLM 2 answers 65% of the time during real-world question evaluations.

Singhal et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark