中文医疗大模型的近况

LLM

FROM BEGINNER TO EXPERT: MODELING MEDICAL KNOWLEDGE INTO GENERAL LLMS

未开源

提出了三阶段训练方法：

1. 医疗领域Post- Training
2. 通用QA 微调
3. 通过C-play 增强下游场景任务。
   第一阶段训练数据：
   

Taiyi: A Bilingual Fine-Tuned Large Language Model for Diverse Biomedical Tasks. 2023

提出了两阶段sft，差不多是大量的低质量多样性预料学习领域知识，医疗数据sft。
开源了数据集清单。

AlpaCare: Instruction-tuned Large Language Models for Medical Application.

提出了类似于self-instruct的医疗数据生成方法，并开源了52k sft data.通过Rouge-L去重。

BianQue: Balancing the Questioning and Suggestion Ability of Health LLMs with Multi-turn Health Conversations Polished by ChatGPT

主张多轮问询CoQ，通过好大夫开源的问答数据，清洗了一遍，然后用ChatGPT润色。

Qilin-Med: Multi-stage Knowledge Injection Advanced Medical Large Language Model

分为三阶段：CPT + SFT + DPO ，公开了使用的数据清单。

Zhongjing: Enhancing the Chinese Medical Capabilities of Large Language Model through Expert Feedback and Real-world Multi-turn Dialogue

开源了数据清单，跑通了CPT + SFT + PPO

LVM

LLava-Med

https://arxiv.org/pdf/2306.00890.pdf

Qilin-Med-VL: Towards Chinese Large Vision-Language Model for General Healthcare

https://arxiv.org/pdf/2310.17956.pdf

BioMed-CLIP

https://arxiv.org/pdf/2303.00915.pdf

RadFM

https://arxiv.org/abs/2308.02463

visual-med alpaca

https://cambridgeltl.github.io/visual-med-alpaca/

ophgml

https://arxiv.org/abs/2306.12174