Predicting Software Changes from Desired Behavior Changes

In the past years, we have seen an explosion of Machine Learning techniques applied to software engineering tasks [1, 10, 12]. The vast majority of these approaches is applied to program code, using Large Language Models (LLMs) to predict token sequences in specific contexts [11]. However, the dynamic nature of programs is hardly exploited; on the contrary, interpreting and predicting the semantics of code remains challenging for LLMs.