Exgen: an innovative web-based system approach to automated examination design using retrieval al-augmented generation and large language model/
Camu, Vashti Karmelli V.
Exgen: an innovative web-based system approach to automated examination design using retrieval al-augmented generation and large language model/ Vashti Karmelli V. Camu, Diane Mae M. Corcino, Patrick I. Nacario, Jamie Jasmine D. Saño, and Paul Adrian O. Torres.-- - Manila: Technological University of the Philippines, 2025. - ix, 142pages: 29cm.
Bachelor's thesis
College Of Science.--
Includes bibliographic references and index.
Examinations play a vital role in the education system, enabling the evaluation of
student learning, strengths, and areas for improvement. For creating effective and diverse
exam questions, which is a time-consuming task for educators, this study presents an AI-
powered exam question generation system designed to streamline and enhance the
assessment creation process. The system utilizes an LLaMA 3 language model combined
with ChromaDB for contextual storage and retrieval, enabling the generation of multiple
question types, including multiple-choice, true or false, identification questions, and more.
The backend includes Bloom’s Taxonomy to ensure cognitive depth and variation in
questions. The development process utilized tools such as Visual Studio Code, Figma, Git,
Google Cloud Platform, and SQLite, among others, to ensure users a smooth and optimized
experience. Evaluation results indicate that the EXGEN Web App was evaluated as 3.72,
which is highly acceptable. The system demonstrated high relevance and accuracy in
generating questions, significantly reducing manual effort. In conclusion, the project offers
an efficient and intelligent solution for automated exam creation, aiding educators in
delivering more consistent and effective assessments.
Exam generation
AI system
Cognitive depth
BTH QA 76 / C36 2025
Exgen: an innovative web-based system approach to automated examination design using retrieval al-augmented generation and large language model/ Vashti Karmelli V. Camu, Diane Mae M. Corcino, Patrick I. Nacario, Jamie Jasmine D. Saño, and Paul Adrian O. Torres.-- - Manila: Technological University of the Philippines, 2025. - ix, 142pages: 29cm.
Bachelor's thesis
College Of Science.--
Includes bibliographic references and index.
Examinations play a vital role in the education system, enabling the evaluation of
student learning, strengths, and areas for improvement. For creating effective and diverse
exam questions, which is a time-consuming task for educators, this study presents an AI-
powered exam question generation system designed to streamline and enhance the
assessment creation process. The system utilizes an LLaMA 3 language model combined
with ChromaDB for contextual storage and retrieval, enabling the generation of multiple
question types, including multiple-choice, true or false, identification questions, and more.
The backend includes Bloom’s Taxonomy to ensure cognitive depth and variation in
questions. The development process utilized tools such as Visual Studio Code, Figma, Git,
Google Cloud Platform, and SQLite, among others, to ensure users a smooth and optimized
experience. Evaluation results indicate that the EXGEN Web App was evaluated as 3.72,
which is highly acceptable. The system demonstrated high relevance and accuracy in
generating questions, significantly reducing manual effort. In conclusion, the project offers
an efficient and intelligent solution for automated exam creation, aiding educators in
delivering more consistent and effective assessments.
Exam generation
AI system
Cognitive depth
BTH QA 76 / C36 2025