Scholarlink

Ensuring the reliability of API systems is critical in modern software development. Traditional API testing methods often require significant manual effort and predefined rule-based frameworks, limiting scalability and adaptability. This research presents a novel fine-tuned LLaMA-3 (8B) model optimized for automated API test case generation. This work leverage Low-Rank Adaptation (LoRA) and 4-bit quantization to efficiently fine-tune the model, significantly reducing computational overhead while maintaining performance. The system processes API request structures and autonomously generates high-relevance test cases per input, covering functional, edge, and boundary scenarios. The model is trained using a structured Alpaca-style prompt methodology, ensuring coherence and adaptability across varied API designs. Experiments demonstrate that our approach enhances test coverage, reduces manual effort, and accelerates the API validation pipeline. This work establishes a scalable framework for AI-driven software testing, making it a viable solution for enterprise DevOps, automated quality assurance, and CI/CD workflows.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Efficient Fine-Tuning of LLaMA-3 for Automated API Test Case Generation Using LoRA and 4-Bit Quantization

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • P. K. Tarun Jaikishan,
  • R. Vijayanandh,
  • B. Anandhu,
  • R. Anuradha

摘要

Ensuring the reliability of API systems is critical in modern software development. Traditional API testing methods often require significant manual effort and predefined rule-based frameworks, limiting scalability and adaptability. This research presents a novel fine-tuned LLaMA-3 (8B) model optimized for automated API test case generation. This work leverage Low-Rank Adaptation (LoRA) and 4-bit quantization to efficiently fine-tune the model, significantly reducing computational overhead while maintaining performance. The system processes API request structures and autonomously generates high-relevance test cases per input, covering functional, edge, and boundary scenarios. The model is trained using a structured Alpaca-style prompt methodology, ensuring coherence and adaptability across varied API designs. Experiments demonstrate that our approach enhances test coverage, reduces manual effort, and accelerates the API validation pipeline. This work establishes a scalable framework for AI-driven software testing, making it a viable solution for enterprise DevOps, automated quality assurance, and CI/CD workflows.