CogniLearn: Integrating AI-Powered Insights for Class 10 Syllabus

Abstract

Cognilearn is an AI-powered educational platform designed to provide high-quality and examination-focused answers for class 10 CBSE exams. The project aims to fine-tune and train AI models to deliver responses that align with board examination standards for Science, Social Studies, and English courses. The study uses two base Language Models namely, Llama2 7B and Qwen $\mathbf{0. 5 B}$. These pre-trained language models are further fine tuned with custom datasets for the question-answering task. Other fine tuning techniques such as LoRA and QLoRa are employed to enhance the capability of the Language Models. The project incorporated Retrieval-Augmented Generation which enables feeding of textbooks and question papers to improve the answer accuracy and relevance. Finally, Prompt Engineering ensures that the retrieved answers are presented appropriately, adhering to standard guidelines. The entire approach ensures that responses are contextually enriched to meet the specific needs of the board examination preparation. The study includes experiments that analyze how effectively these Language Models are used for question-answering tasks in the context of CBSE Class 10 Examination preparation.

Ähnliche Arbeiten