Internistai

Models by this creator

🔄

base-7b-v0.2

internistai

Total Score

56

The base-7b-v0.2 model from Internist.ai is a medical domain large language model trained by medical doctors to demonstrate the benefits of a physician-in-the-loop approach. The training data was carefully curated by medical doctors to ensure clinical relevance and required quality for clinical practice. This 7 billion parameter model is the first to score above the 60% pass threshold on the MedQA (USMLE) benchmark, outperforming similar sized models across most medical evaluations. The model was developed by a team at UCLouvain and Cliniques Universitaires Saint-Luc, and is licensed under the APACHE 2.0 LICENSE. It was continued pretraining from the Mistral-7B-v0.1 model, and has a context length of 4096 tokens with a knowledge cutoff of October 2023. Similar models include Meditron-7B from the EPFL LLM team, which is a 7 billion parameter model adapted from Llama-2-7B through continued pretraining on a medical corpus, and Llama3-OpenBioLLM-70B from Saama AI Labs, a 70 billion parameter open-source biomedical language model. Model Inputs and Outputs Inputs The model accepts text-only data as input. Outputs The model generates text output. Capabilities The base-7b-v0.2 model was trained specifically for the medical domain, demonstrating strong performance on medical question answering and reasoning tasks. It can be used to assist medical professionals with tasks like clinical decision support, disease information query, and general health information lookup. What Can I Use It For? The base-7b-v0.2 model can be a useful tool for medical professionals as an AI assistant, helping with clinical documentation and decision-making. However, the model was not trained for specific safety considerations, so it is not recommended for use by non-professionals who may not be able to identify errors in the model's outputs. Things to Try Researchers and developers could explore fine-tuning or instruction-tuning the base-7b-v0.2 model for more specialized medical tasks and applications. The model could also be used as a starting point for further pretraining on larger medical corpora to create even more capable models tailored for healthcare. Careful evaluation and testing would be essential to ensure the model's outputs are accurate, safe, and appropriate for real-world medical use.

Read more

Updated 5/30/2024