Summary of A Comparative Study Of Zero-shot Inference with Large Language Models and Supervised Modeling in Breast Cancer Pathology Classification, by Madhumita Sushil et al.
A comparative study of zero-shot inference with large language models and supervised modeling in breast cancer pathology classification
by Madhumita Sushil, Travis Zack, Divneet Mandair, Zhiwei Zheng, Ahmed Wali, Yan-Ning Yu, Yuwei Quan, Atul J. Butte
First submitted to arxiv on: 25 Jan 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
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 Medium Difficulty summary: Recent advances in large language models (LLMs) have sparked interest in their potential to reduce the need for large-scale data annotations in information extraction from clinical notes. This study explores whether LLMs can achieve this goal by comparing zero-shot classification capability of GPT-4 and GPT-3.5 with supervised classification performance of various model architectures, including random forests, LSTM networks, and UCSF-BERT. The results show that the GPT-4 model outperforms or matches the best supervised models on 13 tasks, demonstrating its potential to speed up clinical NLP studies by reducing data labeling requirements. However, simpler supervised models can still provide comparable results if large annotated datasets are available. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty summary: This study looks at how well big language models can help us extract important information from medical records without needing a lot of labeled data. The researchers tested these models on 13 different tasks and found that one model, GPT-4, did almost as well as the best trained models. This means we might be able to get similar results with less work if we use these language models instead of collecting a huge amount of labeled data. This could make it easier to do studies using artificial intelligence and medical records. |
Keywords
* Artificial intelligence * Bert * Classification * Data labeling * Gpt * Lstm * Nlp * Supervised * Zero shot
