UltraMedical: Building Specialized Generalists in Biomedicine

TL;DR

This paper introduces UltraMedical, a collection of high-quality biomedical datasets used to fine-tune Llama-3 based models, significantly advancing specialized medical language understanding and addressing privacy concerns in biomedical AI.

Contribution

The paper provides the UltraMedical datasets and demonstrates their effectiveness in fine-tuning biomedical LLMs, improving performance on medical benchmarks and developing reward models for preference learning.

Findings

Fine-tuned models show state-of-the-art performance on medical benchmarks.

UltraMedical datasets enable effective preference learning in biomedical LLMs.

Reward models improve online preference learning in biomedical AI.

Abstract

Large Language Models (LLMs) have demonstrated remarkable capabilities across various domains and are moving towards more specialized areas. Recent advanced proprietary models such as GPT-4 and Gemini have achieved significant advancements in biomedicine, which have also raised privacy and security challenges. The construction of specialized generalists hinges largely on high-quality datasets, enhanced by techniques like supervised fine-tuning and reinforcement learning from human or AI feedback, and direct preference optimization. However, these leading technologies (e.g., preference learning) are still significantly limited in the open source community due to the scarcity of specialized data. In this paper, we present the UltraMedical collections, which consist of high-quality manual and synthetic datasets in the biomedicine domain, featuring preference annotations across multiple advanced LLMs. By utilizing these datasets, we fine-tune a suite of specialized medical models based on Llama-3 series, demonstrating breathtaking capabilities across various medical benchmarks. Moreover, we develop powerful reward models skilled in biomedical and general reward benchmark, enhancing further online preference learning within the biomedical LLM community. Datasets and models are available at https://github.com/TsinghuaC3I/UltraMedical