Skip to content

Resources and codes for SIGIR 2024 paper "Beyond Accuracy: Investigating Error Types in GPT-4 Responses to USMLE Questions"

License

Notifications You must be signed in to change notification settings

roysoumya/usmle-gpt4-error-taxonomy

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Description

This contains the "GPT-4 USMLE Error Dataset" (G4UE) that is accepted as a full paper to the SIGIR 2024 "Resource and Reproducibility Track". The paper is titled "Beyond Accuracy: Investigating Error Types in GPT-4's Responses to USMLE Questions".

The resources are present under the "data" directory.

The semrep annotations consisting of UMLS concepts and semantic predications are present under "semrep" directory.

The codebase for setting up the Potato annotation platform to replicate our span-labeling annotation setup in present under "prolific-potato-annotation" directory.

The "notebooks" folder contains Jupyter notebooks for easy exploration of proposed dataset and annotations.

Resources Overview

Overview of resources created in the paper

MedQA Dataset

The MedQA dataset is not a contribution of this work. It is obtained from the Google Drive link mentioned on the MedQA "Papers with Code" homepage.

In this work, we contribute only the error taxonomy, error type annotations to the MedQA dataset as well as the long form GPT-4 responses to a large portion of the MedQA training dataset.

About

Resources and codes for SIGIR 2024 paper "Beyond Accuracy: Investigating Error Types in GPT-4 Responses to USMLE Questions"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published