Enhancing Solution Diversity in Arithmetic Problems using Fine-Tuned AI Language Model

After the emergence of large language models (LLMs), numerous studies have incorporated LLM into educational contexts. Nevertheless, due to the suboptimal performance of LLM in mathematical computations, the application of math education with the aid of LLM is limited. Meanwhile, as the educational experts indicate that, the students can learn better if diverse solutions are provided. Thus, we propose a methodology that simplifies the LLM to learn through instructional abstraction. By adopting this methodology, the solution diversity is greatly, automatically, and most importantly, accurately generated.