Roma
Via della Vite 41, 00187
+39 06 772 50 136
+39 06 770 70 449
Rende
Rende (CS)
Corso Italia 215, 87036

Language switcher

PC Cube ICST

PC Cube at ICST 2025

We are excited to announce that we have been invited to present our work CUBETESTERAI: Automated JUnit Test Generation using the LLaMA Model at the 18th IEEE International Conference on Software Testing, Verification and Validation (ICST) 2025, which will be held from March 31st to April 4th in Naples.

ICST is one of the leading international conferences dedicated to software testing, a reference event for researchers and practitioners to share the latest innovations in software verification and validation.

CUBETESTERAI: AI for Software Testing

CUBETESTERAI is an advanced solution that leverages the LLaMA 3.2 (Large Language Model Architecture) artificial intelligence model to automate the generation of JUnit tests for Spring Boot-based Java applications, reducing manual work for developers and improving code coverage.

Key features:

  • Automation of JUnit tests directly from the source code, eliminating the need for manual writing.
  • Integration with CI/CD, thanks to the use of GitLab and Docker, to ensure constant and up-to-date tests.
  • Intuitive web interface for simplified test management, accessible even to less experienced developers.
  • Cloud execution via RunPod, ensuring high performance and data protection.
  • Advanced code optimization thanks to iterative refinement that improves quality and reduces errors in generated tests.
  • Testing of private methods without violating the encapsulation principle, allowing for broader code coverage.

The system also includes a continuous feedback mechanism that allows you to refine the generated tests, progressively improving the quality of the automated verification.

Innovation and collaboration are the basis of our project

CUBETESTERAI was born from a thesis developed by Shivam Kumar during the PC Cube AI Coding Camp. The goal was to improve the efficiency of automatic software test generation and provide a reliable solution for developers.

The project took shape thanks to the work of Shivam Kumar himself, Alireza Alipourfaz (mentor during Shivam’s AI Coding Camp), our CTO Pietro Roselli Lorenzini and a close collaboration with Prof. Daniele Gorla of the Department of Computer Science of La Sapienza University of Rome.
After the conclusion of the thesis, considering the excellent results and the interest encountered, the team continued to work on CUBETESTERAI, refining its functionalities and transforming it into an innovative product ready to be applied in the industrial field.
After a rigorous peer review process, our work was selected for the conference in the Industry Track, confirming the solidity of the research and its concrete impact on the sector.

A comparison with existing solutions

CUBETESTERAI stands out for its superior performance compared to other test automation solutions. Our tests have shown that:

  • It achieves code coverage of up to 97%, far superior to alternatives such as ChatGPT, EvoSuite and ChatUniTest documented in the relevant scientific literature.
  • It produces more correct tests with less manual intervention, reducing the time required for software validation.
  • Iterative optimization allows for higher code quality, identifying and eliminating errors with greater precision.

These results demonstrate the concrete value of CUBETESTERAI for the software testing industry, making it a reliable tool for companies and developers.

Looking to the future of AI in software testing

Acceptance at ICST 2025 is a testament to the value of our research and the innovation that PC Cube brings to the industry. But our commitment doesn’t stop there.

We are already working to extend CUBETESTERAI to other programming languages, further improve CI/CD integration, and make test generation even more efficient. Our goal is to make this technology accessible to a wider audience, improving the quality and reliability of software globally.

We also plan to explore integration with additional AI tools to enhance automated code analysis and further optimize the testing process.

We look forward to showcasing CUBETESTERAI at ICST 2025 and engaging with the scientific and industrial community to continue innovating in the field of software testing.

Related posts