Abstract

Developing test oracles can be inefficient - developer generative oracles are time-intensive and thus costly while automatic oracle generation in the form of regression or exception oracles assumes that the underlying code is correct. To mitigate the high cost of testing oracles, the SEER tool was developed to predict test outcomes without needing assertion statements. The creators of SEER introduced the tool with an overall accuracy of 93%, precision of 86%, recall of 94%, and an F1 score of 90%. If these results are replicable on new data with perturbations, i.e. SEER is generalizable and robust, the model would represent a significant advancement in the field of automated testing. Consequently, we conducted a comprehensive reproduction of SEER and attempted to verify the model’s results on a new dataset.

Kasra Lekan

Master’s Student

My research interests include natural language processing, human-AI interaction, and modelling complex systems.