Fusion probability large language model test case generation method and system

Abstract

The present application relates to the field of test case generation, and provides a large language model test case generation method and system fusing probability, which comprises obtaining candidate probability and candidate probability parameter of training assertion, constructing training assertion feature vector through candidate probability parameter and uncertainty mark; obtaining assertion correctness parameter, multiple benchmark models respectively performing multi-level correct probability prediction on the training assertion feature vector to obtain comprehensive prediction probability, and obtaining a judgment model according to the assertion correctness parameter and the comprehensive prediction probability; generating an initial assertion according to assertion generation constraints, calculating the mode coverage number and Shannon entropy of the initial assertion, judging the coverage degree of the initial assertion, and obtaining a high-coverage-degree assertion; calculating the candidate probability parameter and the uncertainty mark of the high-coverage-degree assertion, constructing an assertion feature vector, inputting the assertion feature vector into the judgment model, and obtaining a judgment label; and then judging and rewriting to obtain a target assertion, so as to generate a test case.

Images