LiLinaamari/Llama3-OpenBioLLM-8B
OpenBioLLM-8B is an 8 billion parameter biomedical large language model developed by Saama AI Labs, fine-tuned from Meta-Llama-3-8B. It specializes in understanding and generating text for medical and life sciences, leveraging a custom diverse medical instruction dataset and Direct Preference Optimization (DPO). This model demonstrates superior performance on biomedical benchmarks compared to other open-source models of similar scale and even larger proprietary models like GPT-3.5 and Meditron-70B.
Loading preview...
OpenBioLLM-8B: A Specialized Biomedical LLM
OpenBioLLM-8B, developed by Saama AI Labs, is an 8 billion parameter language model specifically designed for the biomedical domain. Built upon the Meta-Llama-3-8B architecture, it has been fine-tuned using a custom diverse medical instruction dataset and Direct Preference Optimization (DPO) techniques, including the berkeley-nest/Nectar ranking dataset.
Key Capabilities
- Biomedical Specialization: Tailored for the unique language and knowledge requirements of medical and life sciences.
- Superior Performance: Outperforms other open-source biomedical models of similar scale and shows better results than larger models like GPT-3.5 and Meditron-70B on various biomedical benchmarks.
- Advanced Training: Incorporates DPO for aligning with biomedical application preferences.
Benchmark Highlights
OpenBioLLM-8B achieves an average score of 72.50% across 9 diverse biomedical datasets, demonstrating strong performance in tasks such as Clinical KG, Medical Genetics, and PubMedQA. It surpasses Gemini-1.0, GPT-3.5 Turbo, Meditron-70B, and Mistral-7B-v0.1 in overall average performance within this domain.
Use Cases
- Summarizing clinical notes and EHR data.
- Answering medical questions with domain-specific accuracy.
- Performing clinical entity recognition (diseases, symptoms, medications).
- Extracting biomarkers.
- Biomedical classification tasks (disease prediction, document categorization).
- De-identification of personally identifiable information (PII) from medical records.
Important Advisory
This model is intended for research, development, and exploratory applications only. It should not be used for direct patient care, clinical decision support, or professional medical purposes due to potential inaccuracies or biases. Always consult a qualified healthcare provider for personal medical needs.