The script used for Short answer scoring with GPT-4.
GPT_inference.ipynb contains the script for student answer scoring.
Before running anything, be sure to adjust the following variables as needed. All prompts used for experiments are under folder prompt/. Output will be saved in folder gpt-4-output/.
| Variable | Description | Options |
|---|---|---|
question |
Name of the dataset | Q1 |
model |
GPT model that will be called | Models available by OpenAI API |
exp_name |
Experiment to be conducted (corresponds to the experiment names in the paper) | base(score_rationale_onequery), add_possible_answers, add_scored_examples, add_both, score_only, rationale_score_twoqueries, rationale_score_onequery |
The paper will appear at L@S 2024. If you found this script useful please cite our work as:
@inproceedings{jiang2024short,
author = {Jiang, Lan and Bosch, Nigel},
title = {Short answer scoring with GPT-4},
booktitle = {In Proceedings of the Eleventh ACM Conference on Learning @ Scale (L@S ’24)},
year = {2024},
}