Summary of Fine-tuning Large Language Model (llm) Artificial Intelligence Chatbots in Ophthalmology and Llm-based Evaluation Using Gpt-4, by Ting Fang Tan et al.
Fine-tuning Large Language Model (LLM) Artificial Intelligence Chatbots in Ophthalmology and LLM-based evaluation using GPT-4
by Ting Fang Tan, Kabilan Elangovan, Liyuan Jin, Yao Jie, Li Yong, Joshua Lim, Stanley Poh, Wei Yan Ng, Daniel Lim, Yuhe Ke, Nan Liu, Daniel Shu Wei Ting
First submitted to arxiv on: 15 Feb 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed study assesses the alignment of GPT-4-based evaluation with human clinician experts for evaluating responses generated by fine-tuned language models (LLMs) chatbots to ophthalmology-related patient queries. Five LLMs are fine-tuned and tested on a dataset of 400 questions, and their responses are evaluated using a customized clinical rubric. The results show that GPT-3.5 scores the highest, followed by LLAMA2-13b, LLAMA2-13b-chat, LLAMA2-7b-Chat, and LLAMA2-7b. The GPT-4 evaluation demonstrates significant agreement with human clinician rankings, highlighting its potential to streamline clinical evaluation of LLM chatbot responses. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study checks if a computer program (GPT-4) is good at judging answers from another computer program (LLMs) that help doctors answer patient questions. The programs were tested on 400 questions and the answers were looked at using special rules for what makes an answer good or bad. The results show that one program, GPT-3.5, does a great job of agreeing with what human doctors think is a good answer. This could be useful in helping to make sure that computer programs are giving good answers when they’re used in healthcare. |
Keywords
» Artificial intelligence » Alignment » Gpt
