TPSA 2027

While decades of research in program logics, abstract interpretation, and separation logic provide a strong foundation for automated static analysis, new challenges have arisen in recent years, spurring accelerated innovation in both the underlying theory and practical implementation of program analysis systems. In particular, the rapid adoption of generative AI tools in programming workflows means that software is being created at an unprecedented pace, and with less human scrutiny than ever before. It has been widely reported that AI-generated code contains bugs that are difficult to catch through informal inspection and code review, meaning that automated program analysis with formal guarantees remains highly relevant.

For the past four years, the Theory and Practice of Static Analysis (TPSA) workshop has provided a venue for researchers to share cutting edge work, which promises to improve the quality of production software. Earlier iterations of TPSA focused heavily on the foundations of Incorrectness reasoning and formal techniques to guarantee that reported bugs are true positives. Foundational ideas presented at TPSA have been adopted at Meta and Bloomberg. In the current iteration, we continue to welcome submissions on logical foundations, and also plan to expand the scope to feature both AI-driven program analysis and case studies where assurances are provided for AI generated code.

As we enter a new era in AI-driven software development, program analysis is becoming increasingly relevant. But due to the speed at which AI programming has been adopted, the static analysis community is still working to catch up to the status quo. Since it is very hard to implement and validate new program analysis systems, TPSA provides an important venue for sharing early stage ideas and emerging trends that are not yet ready for publication in conferences like POPL, VMCAI, CAV, or SAS.

Call for Presentations

We invite the submission of talk proposals in topics related to both the mathematical foundations and practical implementations of static analysis, with a particular interest for the integration of the tools in AI pipelines. This workshop will not have formal proceedings, so talks covering in-progress or already published work are welcome. Since analysis tools and algorithms are difficult to implement, we also welcome speculative presentations about techniques that are not yet validated. The topics in scope include, but are not limited to:

• Logical foundations for analysis algorithms (e.g. program logics, abstract interpretation, separation logic, etc)
• Emerging problems and use cases for static analysis (with or without proposed solutions)
• Prototype analysis tools
• Integration of static analysis with AI agents
• Use of static analysis for AI training
• Incorrectness, under-approximation, and bug-finding
• Analysis with computational effects (e.g., probabilistic, quantum, or concurrent programming)
• Industrial experience reports

Submissions should be in the form of extended abstracts and must not exceed three pages (excluding references) in the SIGPLAN two-column format.