I work in software reliability (bug finding through dynamic program analysis) which is a related domain of this research.

Most of these machine learning based software engineering research tools are based on unrealistic scenarios, full of over-promises and very little to deliver in real life.