CodeFuse-TestGPT-7B is a 7 billion parameter large language model developed by Ant Group, built upon the CodeLlama-7B foundation. It is specifically fine-tuned for software quality assurance tasks, excelling in multi-language test case generation for Java, Python, and JavaScript, and test case assertion completion for Java. The model demonstrates superior performance in these specialized tasks compared to general-purpose LLMs, making it ideal for automating and enhancing software testing workflows.
Loading preview...
Overview
CodeFuse-TestGPT-7B is a 7-billion parameter model developed by Ant Group, specifically designed for software quality domains. It is built on the CodeLlama-7B architecture and has been extensively fine-tuned for critical testing tasks.
Key Capabilities
- Multi-language Test Case Generation: Generates test cases for Java, Python, and JavaScript, with future support planned for Go and C++. It addresses limitations of traditional tools by improving readability, scenario completeness, and multi-language support.
- Test Case Assertion Completion: Automatically completes missing assertion statements in existing test cases, particularly for Java. This capability helps improve the quality and effectiveness of regression testing by ensuring tests can detect issues.
Performance Highlights
- Test Case Generation (Pass@1):
- Java: 48.6% (vs. CodeLlama-13B-Instruct 40.54%)
- Python: 35.67% (vs. CodeLlama-13B-Instruct 30.57%)
- JavaScript: 36% (vs. CodeLlama-13B-Instruct 31.7%)
- Assertion Completion (Pass@1):
- Java: 71.1%
Good For
- Developers and QA engineers looking to automate and enhance test case creation.
- Improving the quality of existing test suites by adding missing assertions.
- Projects requiring robust, readable, and comprehensive test cases across multiple programming languages.