Multi-agent large multimodal model (LMM) system and method for generating test cases

Abstract

A system including multiple artificial intelligence (AI) agents generates test cases for a test target in multiple stages. A first AI agent uses a large multimodal model (LMM) to process a first prompt and specifications of the test target to generate respective identifiers of test plans. A second AI agent uses the LMM to process a second prompt, the specifications of the test target, and an output of the first AI agent to generate an ordered list of test steps in each test plan. A third AI agent uses the LMM to process a third prompt and an output of the second AI agent to expand the test plans into test cases. The system verifies the output of each of the first AI agent, the second AI agent, and the third AI agent, and generates a test case document according to the test cases.

Images