This repository serves as a curated collection of academic papers and resources focused on leveraging Large Language Models (LLMs) for various software testing tasks. It aims to provide researchers and practitioners with a comprehensive overview of the current landscape, emerging trends, and practical applications of LLMs in enhancing software quality and reliability.

How It Works

The collection analyzes LLM applications across the software testing lifecycle, from test case generation (unit, system) and test oracle creation to bug analysis, debugging, and program repair. It categorizes studies based on the LLM perspective, highlighting the prevalence of models like ChatGPT, Codex, and CodeT5, and detailing methodologies such as prompt engineering (zero-shot, few-shot, chain-of-thought) versus model fine-tuning.

Quick Start & Requirements

This repository is a collection of research papers and does not have direct installation or execution commands. It serves as a knowledge base.

Highlighted Details

• Comprehensive analysis of 102 papers on LLMs in software testing.
• Categorization of LLM usage from both software testing and LLM perspectives.
• Detailed breakdown of LLM techniques: prompt engineering (zero-shot, few-shot) and fine-tuning.
• Extensive lists of papers covering unit test generation, test oracle generation, system test input generation, bug analysis, debugging, and program repair.

Maintenance & Community

The project is under active development, with recent papers accepted at ICSE 2024 and publications in ICSE 2023. Users are encouraged to STAR and WATCH the repository for updates.

Licensing & Compatibility

The repository itself is a collection of links to academic papers. The licensing of the individual papers would need to be checked on their respective publication platforms.

Limitations & Caveats

This repository is a survey and does not provide executable code or tools. The effectiveness and practical implementation of the discussed LLM techniques would depend on the specific LLM used, the quality of prompts, and the nature of the software being tested.