The invention relates to the technical field of
artificial intelligence, in particular to a
model test management method and device,
electronic equipment and a storage medium, and the method comprises the steps: constructing a
test case set containing standard questions and a plurality of similar questions corresponding to each standard question, and inputting the
test case set into a preset to-be-tested dialogue model, obtaining an answer text and an answer
confidence score corresponding to each similar question, and calculating a single
score of each similar question; and based on the single
score of each similar question, calculating to obtain an overall
score of a preset to-be-tested dialogue model, and when the overall score of the preset to-be-tested dialogue model meets a preset condition, taking a current version of the preset to-be-tested dialogue model as a target online version. Therefore, the problems of low efficiency, subjective evaluation and the like caused by dependence on manpower in an existing AI
model test are solved, the performance of the AI model can be automatically, efficiently and objectively evaluated, the test labor cost is remarkably reduced, and the quantifiable and comparable model
iteration process is ensured.