Large Language Models

Fillmore examines the integration of large language models into cardio-oncology to improve the monitoring of cardiotoxicity and the precision of patient treatment planning. His work proposes a scalable framework for customizing these models with institutional data and clinical guidelines within electronic health records. In evaluating several architectures, he found that GPT-4 demonstrated superior performance compared to Gemini 1.5 and LLaMA-3-70B for clinical applications in this specialty.