arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.12182cs.CLcs.AI
收藏

Aquila-Med LLM:开创全流程开源医学语言模型先河

Aqulia-Med LLM: Pioneering Full-Process Open-Source Medical Language Models

Beijing Academy of Artificial Intelligence, BAAI(北京人工智能研究院(BAAI));Beijing University of Posts and Telecommunications(北京邮电大学)

Lulu Zhao, Weihao Zeng, Xiaofeng Shi, Hua Zhou, Donglin Hao, Yonghua Lin

AI总结:

针对现有开源大语言模型在医学领域表现不佳的问题,研究团队基于Aquila构建双语医学LLM Aquila-Med,通过多阶段训练与专属数据集实现多医学任务性能提升,并开源相关资源。

摘要:

近期,闭源大语言模型(LLM)与开源社区均取得显著进展,在多个通用领域表现超越人类。但受医学知识复杂性影响,这类模型在医学等特定专业领域的表现仍欠佳,开源社区的相关模型尤为突出。我们提出基于Aquila的双语医学大语言模型Aquila-Med,通过持续预训练、监督微调(SFT)以及人类反馈强化学习(RLHF)应对上述挑战。我们构建了用于持续预训练的大规模中英医学数据集,以及覆盖广泛医学专科的高质量SFT数据集,此外还开发了高质量直接偏好优化(DPO)数据集以进一步对齐模型。Aquila-Med在单轮对话、多轮对话及医学选择题任务中均取得显著结果,验证了我们方法的有效性。我们开源了相关数据集与完整训练流程,为研究界提供宝贵资源。我们的模型与数据集将发布于https://huggingface.co/BAAI/AquilaMed-RL。

英文摘要:

Recently, both closed-source LLMs and open-source communities have made significant strides, outperforming humans in various general domains. However, their performance in specific professional fields such as medicine, especially within the open-source community, remains suboptimal due to the complexity of medical knowledge. We propose Aquila-Med, a bilingual medical LLM based on Aquila, addressing these challenges through continue pre-training, supervised fine-tuning (SFT), and reinforcement learning from human feedback (RLHF). We construct a large-scale Chinese and English medical dataset for continue pre-training and a high-quality SFT dataset, covering extensive medical specialties. Additionally, we develop a high-quality Direct Preference Optimization (DPO) dataset for further alignment. Aquila-Med achieves notable results across single-turn, multi-turn dialogues, and medical multiple-choice questions, demonstrating the effectiveness of our approach. We open-source the datasets and the entire training process, contributing valuable resources to the research community. Our models and datasets will released at https://huggingface.co/BAAI/AquilaMed-RL.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑