Call Graph Pruning
Call graph pruning aims to improve the accuracy of statically-generated call graphs by removing spurious edges, thereby enhancing the precision of program analysis. Recent research focuses on leveraging machine learning, particularly employing transformer-based models and other neural networks, to learn patterns distinguishing true and false call relationships from structural and semantic features extracted from source code and the call graph itself. This work addresses the inherent imprecision of static analysis, leading to more efficient and accurate program analysis tools for tasks such as software testing and debugging. The resulting smaller, more precise call graphs offer significant speed improvements while maintaining the effectiveness of downstream analyses.