Demystifying What Code Summarization Models Learned
This work addresses the challenge of interpreting learned patterns in models for researchers in code analysis and machine learning, offering a new direction for formal analysis, though it is incremental in applying existing formal methods to a specific domain.
The paper tackles the problem of understanding what patterns code summarization models learn by providing the first formal definition of these patterns and a sound algorithm to infer context-free grammars (CFGs) that describe them, applied to models like code2vec and code2seq, revealing that learned patterns are mostly local and syntactic with little semantic meaning.
Study patterns that models have learned has long been a focus of pattern recognition research. Explaining what patterns are discovered from training data, and how patterns are generalized to unseen data are instrumental to understanding and advancing the pattern recognition methods. Unfortunately, the vast majority of the application domains deal with continuous data (i.e. statistical in nature) out of which extracted patterns can not be formally defined. For example, in image classification, there does not exist a principle definition for a label of cat or dog. Even in natural language, the meaning of a word can vary with the context it is surrounded by. Unlike the aforementioned data format, programs are a unique data structure with a well-defined syntax and semantics, which creates a golden opportunity to formalize what models have learned from source code. This paper presents the first formal definition of patterns discovered by code summarization models (i.e. models that predict the name of a method given its body), and gives a sound algorithm to infer a context-free grammar (CFG) that formally describes the learned patterns. We realize our approach in PATIC which produces CFGs for summarizing the patterns discovered by code summarization models. In particular, we pick two prominent instances, code2vec and code2seq, to evaluate PATIC. PATIC shows that the patterns extracted by each model are heavily restricted to local, and syntactic code structures with little to none semantic implication. Based on these findings, we present two example uses of the formal definition of patterns: a new method for evaluating the robustness and a new technique for improving the accuracy of code summarization models. Our work opens up this exciting, new direction of studying what models have learned from source code.