Question Answering On Medqa Usmle

评估指标

Accuracy

评测结果

各个模型在此基准测试上的表现结果

		Paper Title	Repository
Med-Gemini	91.1	Capabilities of Gemini Models in Medicine	-
GPT-4	90.2	Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in Medicine
Med-PaLM 2	85.4	Towards Expert-Level Medical Question Answering with Large Language Models
Med-PaLM 2 (CoT + SC)	83.7	Towards Expert-Level Medical Question Answering with Large Language Models
Med-PaLM 2 (5-shot)	79.7	Towards Expert-Level Medical Question Answering with Large Language Models
MedMobile (3.8B)	75.7	MedMobile: A mobile-sized language model with expert-level clinical capabilities
Meerkat-7B	74.3	Small Language Models Learn Enhanced Reasoning Skills from Medical Textbooks	-
Meerkat-7B (Single)	70.6	Small Language Models Learn Enhanced Reasoning Skills from Medical Textbooks	-
Meditron-70B (CoT + SC)	70.2	MEDITRON-70B: Scaling Medical Pretraining for Large Language Models
Flan-PaLM (540 B)	67.6	Large Language Models Encode Clinical Knowledge
LLAMA-2 (70B SC CoT)	61.5	MEDITRON-70B: Scaling Medical Pretraining for Large Language Models
Shakti-LLM (2.5B)	60.3	SHAKTI: A 2.5 Billion Parameter Small Language Model Optimized for Edge AI and Low-Resource Environments	-
Codex 5-shot CoT	60.2	Can large language models reason about medical questions?
LLAMA-2 (70B)	59.2	MEDITRON-70B: Scaling Medical Pretraining for Large Language Models
VOD (BioLinkBERT)	55.0	Variational Open-Domain Question Answering
BioMedGPT-10B	50.4	BioMedGPT: Open Multimodal Generative Pre-trained Transformer for BioMedicine
PubMedGPT (2.7 B)	50.3	Large Language Models Encode Clinical Knowledge
DRAGON + BioLinkBERT	47.5	Deep Bidirectional Language-Knowledge Graph Pretraining
BioLinkBERT (340 M)	45.1	Large Language Models Encode Clinical Knowledge
GAL 120B (zero-shot)	44.4	Galactica: A Large Language Model for Science

0 of 27 row(s) selected.