Program Comprehension

Program comprehension aims to enable computers to understand the meaning and functionality of source code, facilitating tasks like code summarization, search, and maintenance. Current research heavily utilizes machine learning, particularly large language models (LLMs) and graph neural networks (GNNs) operating on code representations such as abstract syntax trees (ASTs) or token sequences, often incorporating techniques like fuzz testing to improve model training data. These advancements are crucial for improving software development efficiency, reducing maintenance costs, and enhancing the reliability of software systems.