Clinical Camel: An Open Expert-Level Medical Language Model with Dialogue-Based Knowledge Encoding

TL;DR

Clinical Camel is an open, expert-level medical language model fine-tuned from LLaMA-2, achieving state-of-the-art results on medical benchmarks and introducing dialogue-based knowledge encoding for clinical research.

Contribution

This work presents Clinical Camel, a novel open medical LLM with superior benchmark performance and a new dialogue-based knowledge encoding method for clinical data synthesis.

Findings

Outperforms GPT-3.5 on multiple medical benchmarks

Achieves 64.3% on USMLE Sample Exam

Demonstrates capabilities in synthesizing clinical notes

Abstract

We present Clinical Camel, an open large language model (LLM) explicitly tailored for clinical research. Fine-tuned from LLaMA-2 using QLoRA, Clinical Camel achieves state-of-the-art performance across medical benchmarks among openly available medical LLMs. Leveraging efficient single-GPU training, Clinical Camel surpasses GPT-3.5 in five-shot evaluations on all assessed benchmarks, including 64.3% on the USMLE Sample Exam (compared to 58.5% for GPT-3.5), 77.9% on PubMedQA (compared to 60.2%), 60.7% on MedQA (compared to 53.6%), and 54.2% on MedMCQA (compared to 51.0%). In addition to these benchmarks, Clinical Camel demonstrates its broader capabilities, such as synthesizing plausible clinical notes. This work introduces dialogue-based knowledge encoding, a novel method to synthesize conversational data from dense medical texts. While benchmark results are encouraging, extensive and rigorous human evaluation across diverse clinical scenarios is imperative to ascertain safety before implementation. By openly sharing Clinical Camel, we hope to foster transparent and collaborative research, working towards the safe integration of LLMs within the healthcare domain. Significant challenges concerning reliability, bias, and the potential for outdated knowledge persist. Nonetheless, the transparency provided by an open approach reinforces the scientific rigor essential for future clinical applications.