In this on-demand demonstration, GenomOncology’s chief technology officer discusses the use of ChatGPT plugins and knowledge graphs to ground Medical Q&A, or MedQA -- a dataset used to test the performance of models answering medical questions -- and reduce hallucinations in large language models.
In this video, learn more about how:
Medical Q&A is a dataset used to test the performance of models answering medical questions.
LLMs are prone to hallucinations due to a limited understanding, data limitations, and no real-time interaction capability.
Grounding strategies include pre-training and fine-tuning, prompt engineering, and similarity search to provide more context and improve the model’s performance.
Knowledge graphs can augment LLM’s performance by providing access to databases of specialized fields like genomics.