Deleted Journal, Journal Year: 2025, Volume and Issue: 28(1)

Published: May 11, 2025

Abstract This study investigates the efficiency of large language model (LLM) application programming interfaces (APIs)—specifically GPT-4 and Llama-3—as AI tutors for undergraduate Geotechnical Engineering education. As educational needs in specialised fields like become increasingly complex, innovative teaching tools that provide personalised learning experiences are essential. Unlike previous studies on AI-driven education, our research uniquely focuses assessing role retrieval-augmented generation (RAG) improving accuracy LLM-generated solutions to problems. A dataset 391 questions from related textbook written by Das Sobhan (Das B, K. Principles engineering, Eight Edition. In: Cengage Learning. 2014) was used evaluation, with sourced textbook’s manual. Performance benchmarking focused 20 challenging previously identified Chen et al. (Chen Geotechnics 4:470–498, 2024) as problematic Zero Shot tasks. API support demonstrated superior accuracy, achieving rates 95% at a temperature setting 0.1, 82.5% 0.5, 60% 1. In comparison, Llama-3 achieved an 25% tasks 45% 0.1. The findings highlight GPT-4’s potential tutor education while demonstrating need domain-specific optimisation advanced formula integration techniques. contributes ongoing discourse providing empirical evidence supporting deployment LLMs personalised, adaptive aids engineering disciplines. Future work should explore optimised strategies, expanded domain knowledge bases, long-term student outcomes.

Language: Английский