How Large AI Models Transform Software Testing: From Code Review to Automated Test Cases

1. Concept of Large Models

Large models refer to machine‑learning models with billions or even trillions of parameters, typically built from deep neural networks. Their purpose is to increase expressive power and prediction performance, enabling them to handle complex tasks and data across domains such as natural language processing, computer vision, speech recognition, and recommendation systems.

ChatGPT describes large models as deep neural networks trained on massive data, whose scale leads to emergent intelligence that resembles human reasoning.

2. Applying Large Models to Software Testing

1. Test Left‑Shift

1.1 Code Review

Our company offers large‑model tools such as JoyCoder, Autobot, and aichat.jd. For example, JoyCoder can explain selected code, helping testers quickly understand it.

1.2 Generate Pre‑Test Cases

When product requirements are simple and clear, JoyCoder can directly generate preliminary test cases.

2. Testing Process

2.1 Code Security Check

Using JoyCoder → “神医安全检查”, you can preliminarily determine whether the code contains security vulnerabilities.

2.2 Code Performance Check

Select a segment of code, then type in JoyCoder’s dialog: “检测我选中的这段代码的性能” and press Enter to evaluate performance.

3. Conclusion and Outlook

Feeding a large model with comprehensive product documentation can shift its reasoning toward product‑specific business logic; once mature, it could design about 90 % of manual test cases, freeing the saved time for writing automated test scripts and other tasks.