ZK-GenMed: A Zero-shot Knowledge Generative Medical Large Language Model
Abstract: Advancements in Natural Language Processing (NLP) have led to the development of Large Language Models (LLMs), which have demonstrated remarkable capabilities in various tasks, domains, and settings. These models have demonstrated efficacy in various training and evaluation scenarios, including zero-shot learning and instruction settings. They have been effectively applied to reasoning, summarizing, and answering questions. Moreover, LLMs have been used in various industries, including the medical profession, where they have been used for jobs requiring accuracy, such as answering questions. However, much research hasn't been done on LLMs' potential for resolving medical questions in a zero-shot manner. To close this knowledge gap, we provide a novel framework called ZK-GenMed, which uses LLMs' advantages to produce the information needed to answer medical questions in a zero-shot scenario. This framework combines the generated knowledge with ranking strategies to extract relevant information, meaningfully enabling the model to answer medical questions. Experimental results demonstrate significant improvements, with marginal gains of over 10\% on various datasets, highlighting the potential of ZK-GenMed for medical question-answering applications.
Paper Type: Long
Research Area: NLP Applications
Research Area Keywords: LLM, zero-shot, medical question-answering, knowledge Generation, prompting
Contribution Types: NLP engineering experiment
Languages Studied: English
Submission Number: 5527
Loading