Solution generation for university math problems using large language models

• المؤلف الرئيسي: Wirja, Louis
• مؤلفون آخرون: Owen Noel Newton Fernando
• التنسيق: Final Year Project
• اللغة: English
• منشور في: Nanyang Technological University 2024
• الموضوعات: Computer and Information Science; Large language models; Generative AI; Deep learning; Transformers; Machine learning; Prompt engineering
• الوصول للمادة أونلاين: https://hdl.handle.net/10356/175331

This study assesses the capabilities of cutting-edge Large Language Models (LLMs) including GPT-3.5 Turbo, GPT-4, and Gemini Pro in solving university-level math problems, with a focus on enhancing both accuracy and comprehensive explanation generation to aid in mathematical reasoning education. Through rigorous evaluation on a curated problem set spanning university calculus topics, we explore zero-shot and few-shot learning scenarios, measuring performance via accuracy and semantic similarity. Additionally, an ensemble model combining GPT-3.5 Turbo and Gemini Pro shows improved efficacy compared to individual components. By combining precise solutions with clear, step-by-step explanations, our study aims to provide students with vital tools for learning complex concepts and developing mathematical intuition