Generating Test Cases with Large Language Model: Frameworks and Challenges

Software testing is a fundamental stage in the Software Development Life Cycle (SDLC), indispensable for detecting errors and lowering overall maintenance costs. Automated test generation tools aim to simplify test creation by producing effective, bug-revealing test cases. Recently, Large Language Models (LLMs) trained on code have been applied to unit test generation, showing promising improvements in readability and coverage. Various frameworks have been discussed in this paper, each using LLMs to generate test cases through two different approaches: first, fine-tuning LLMs to generate test cases, and second, prompting LLMs to create test cases. In this study, we review and analyze these frameworks and outline the steps they follow to generate test cases. In addition, we compare their strengths, limitations, highlighting both their similarities and differences to provide insights into current trends and future directions in LLM-based test case generation.