chambul/MediSimplifier-OpenBioLLM-merged

MediSimplifier-OpenBioLLM-merged: Medical Text Simplification

This model, developed by chambul, is an 8 billion parameter language model based on the Meta Llama 3 architecture, specifically fine-tuned from aaditya/Llama3-OpenBioLLM-8B. Its core purpose is the simplification of medical text.

Key Capabilities

• Medical Text Simplification: The model excels at rephrasing complex medical jargon into more understandable language, making medical information accessible to a broader audience.
• LoRA Fine-tuning: It incorporates a merged LoRA adapter (r=32, all_attn, rsLoRA) applied to the Q/K/V/O target modules, trained over 3 epochs on the GuyDor007/medisimplifier-dataset.
• Performance: Achieved a ROUGE-L score of 0.6638 on a 1,001-sample test split, indicating its effectiveness in generating simplified text while retaining key information.
• Llama 3 Foundation: Benefits from the robust capabilities and architecture of the Meta Llama 3 base model.

Good For

• Healthcare Applications: Ideal for tools that need to simplify patient-facing medical documents, health information portals, or clinical notes.
• Educational Resources: Can be used to create more accessible educational materials for medical students or the general public.
• Information Accessibility: Enhancing the readability of scientific papers, research summaries, or pharmaceutical information for non-experts.