# Model Fusion Optimization

Named entity recognition model optimized with model fusion technology, F1 score improved by 3.25 points to 0.6601

A merged model based on multiple 8B-parameter Llama-3.1 models, optimized using the model_stock method

An open-source large language model for the medical domain based on the Mistral architecture, further pre-trained on PubMed Central texts

BioMistral is an open-source large language model optimized for the medical domain based on the Mistral architecture, further pre-trained on PubMed Central open-access text data, supporting multilingual medical question-answering tasks.

BioMistral is an open-source medical domain model suite based on the Mistral architecture, further pretrained using PubMed Central open-access text data.

This model is created by merging OpenHermes-2.5-Mistral-7B and Intel's neural-chat-7b-v3-2 using the ties merging method, focusing on text generation tasks.