用巴西临床指南训练大模型，提升医疗问答准确率。
Teaching LLMs Brazilian Healthcare: Injecting Knowledge from Official Clinical Guidelines
- 基于178份指南生成7000万条合成数据，持续预训练并强化学习优化。
- 在自建基准上达到83.9%准确率，超越多个主流大模型。
- 专为巴西葡语临床场景设计，适合医疗AI研究者使用。
巴西统一医疗体系（SUS）依赖官方临床指南，为超过2亿公民定义诊断标准、治疗方案、用药剂量和监测流程。然而现有大语言模型在该领域表现不佳，且缺乏基于巴西葡萄牙语指南的评估基准。本文将Qwen2.5-14B-Instruct适配至巴西临床领域，从178份官方指南（约540万词元）中，利用四类生成模型生成约7000万词元的合成数据，涵盖重述、维基风格文章及问答对三种格式。随后采用持续预训练结合组相对策略优化（GRPO）。提出HealthBench-BR（含1780个平衡真/假临床断言）与PCDT-QA（890个开放问答，由LLM评分），最佳模型在HealthBench-BR上达83.9%，在PCDT-QA上达85.4%，超越GPT-5.2、Claude Sonnet 4.6、Gemini 3.1 Pro及Google AI Overview的网页增强RAG，尽管仅含140亿参数。消融实验表明生成多样性与强化学习对性能提升至关重要。所有数据集、基准与模型权重均已开源，支持巴西葡语临床NLP研究可复现性。代码、数据与模型权重见https://github.com/hugoabonizio/clinical-protocols-br。
原文摘要 · Abstract (English)
Brazil's Unified Health System (SUS) relies on official clinical guidelines that define diagnostic criteria, treatments, dosages, and monitoring procedures for over 200 million citizens. Yet current LLMs perform poorly on this guideline-specific knowledge, and no benchmark evaluates clinical recall grounded in Brazilian Portuguese protocols. We address this gap by adapting Qwen2.5-14B-Instruct to the Brazilian clinical domain. From 178 official guidelines (~5.4M tokens), we generate ~70M tokens of synthetic data in three formats -- rephrases, wiki-style articles, and question-answer pairs -- using four generator LLMs. We then apply continual pre-training followed by Group Relative Policy Optimization (GRPO). We introduce HealthBench-BR, with 1,780 balanced true/false clinical assertions, and PCDT-QA, with 890 open-ended clinical questions scored by an LLM judge. Our best model achieves 83.9% on HealthBench-BR and 85.4% on PCDT-QA, outperforming GPT-5.2, Claude Sonnet 4.6, Gemini 3.1 Pro, and Google AI Overview's web-grounded RAG despite having only 14B parameters. Ablations show that generator diversity and reinforcement learning are critical to these gains. We release all datasets, benchmarks, and model weights to support reproducible clinical NLP research for Brazilian Portuguese. Code, data, and model weights are available at https://github.com/hugoabonizio/clinical-protocols-br
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。