PLDI 2025
Mon 16 - Fri 20 June 2025 Seoul, South Korea
Mon 16 Jun 2025 14:00 - 15:00 at Orchid - SOAP 3 Chair(s): Mukund Raghothaman

Static analysis has long served as a foundation for scalable code analysis and verification by extracting and reasoning about abstract representations of programs. In this talk, we explore how the principles of static analysis, such as those embodied in our SVF framework, can naturally extend to new domains, including the verification of neural networks. We highlight a shared foundation between program analysis and neural network verification, grounded in abstraction, approximation, and sound reasoning over large input spaces. In particular, we present two recent works: an order-leading approach to robustness verification of neural networks and a counterexample-guided incremental verification method. Our initial efforts suggest that static reasoning techniques could offer useful insights for addressing emerging challenges in the quality and reliability of AI systems.

Mon 16 Jun

Displayed time zone: Seoul change

14:00 - 15:20
SOAP 3SOAP at Orchid
Chair(s): Mukund Raghothaman University of Southern California
14:00
60m
Keynote
Static Analysis from Code Graphs to Neural Networks
SOAP
Yulei Sui University of New South Wales
15:00
20m
Talk
Compositional Static Callgraph Reachability Analysis for WhatsApp Android App Health
SOAP
Ákos Hajdu Meta, Roman Lee Meta, Gavin Weng Meta, Nilesh Agrawal Meta, Jérémy Dubreil Unaffiliated
DOI