International Journal of Computer Applications
Foundation of Computer Science (FCS), NY, USA
|
Volume 187 - Issue 21 |
Published: July 2025 |
Authors: Augustine O. Ugbari, Clement Ndeekor, Echebiri Wobidi |
![]() |
Augustine O. Ugbari, Clement Ndeekor, Echebiri Wobidi . Optimizing GPT-4 for Automated Short Answer Grading in Educational Assessments. International Journal of Computer Applications. 187, 21 (July 2025), 32-36. DOI=10.5120/ijca2025925255
@article{ 10.5120/ijca2025925255, author = { Augustine O. Ugbari,Clement Ndeekor,Echebiri Wobidi }, title = { Optimizing GPT-4 for Automated Short Answer Grading in Educational Assessments }, journal = { International Journal of Computer Applications }, year = { 2025 }, volume = { 187 }, number = { 21 }, pages = { 32-36 }, doi = { 10.5120/ijca2025925255 }, publisher = { Foundation of Computer Science (FCS), NY, USA } }
%0 Journal Article %D 2025 %A Augustine O. Ugbari %A Clement Ndeekor %A Echebiri Wobidi %T Optimizing GPT-4 for Automated Short Answer Grading in Educational Assessments%T %J International Journal of Computer Applications %V 187 %N 21 %P 32-36 %R 10.5120/ijca2025925255 %I Foundation of Computer Science (FCS), NY, USA
Automated Short Answer Grading Systems (ASAGS) have witnessed significant advancement with the integration of large language models (LLMs), particularly GPT-4. This paper explores methodologies to optimize GPT-4 for the purpose of grading short answer questions in educational assessments. The focus is on aligning GPT-4’s natural language processing capabilities with human grading rubrics to enhance accuracy, consistency, and fairness. We examine techniques including prompt engineering, rubric-based scoring, and fine-tuning strategies. The research also assesses the model’s performance across various domains, evaluates inter-rater reliability with human graders, and addresses concerns related to bias, explainability, and scalability. This paper proposes a framework that leverages GPT-4 as a co-grader, ensuring human-in-the-loop moderation to improve educational outcomes.