Hugging Face Overview
Overview
FreedomIntelligence/Apollo-7B is an 8.5 billion parameter multilingual medical large language model developed by FreedomIntelligence. It is part of a family of Apollo models ranging from 0.5B to 72B parameters, specifically designed to democratize medical AI globally. The model is trained on the extensive ApolloCorpus and evaluated using the XMedBench, both developed by FreedomIntelligence.
Key Capabilities
- Multilingual Medical Support: Covers English, Chinese, French, Hindi, Spanish, and Arabic for medical contexts.
- Specialized Medical Training: Utilizes a diverse dataset including medical books, guidelines, papers, web content, and QA pairs.
- Comprehensive Evaluation: Benchmarked across multiple medical datasets in various languages, such as MedQA-USMLE, MedMCQA, MMLU-Medical (EN), MedQA-MCMLE, CMMLU-Medical (ZH), Head_qa (ES), Frenchmedmcqa (FR), and MMLU_HI/MMLU_Ara (HI/AR).
- Instruction-tuned: Formatted for user-assistant interactions, making it suitable for conversational medical applications.
Good for
- Medical Question Answering: Providing information based on a wide range of medical texts.
- Multilingual Medical Information Retrieval: Accessing and processing medical knowledge in several languages.
- Research and Development: Serving as a foundation for further specialized medical AI applications, particularly in non-English speaking regions.