As per last year, I wrote a summary of the Computer Aided Verification Conference (CAV2026) I attended in July. This summary usually serves as a basis for a presentation I give when coming back from the conference. It helps me to prioritize the papers to read and summarize, the people to contact and to make sense of what happened.
Highlights We got two papers accepted at SAIV Our tool PyRAT ranked third at VNN-COMP 2026 Our work was featured in a CAV keynote - Formal Explanations of Neural Networks Technical comments Regarding Verification of Neural Networks, formal verification with absolute guarantees will only get you so far. Probabilistic guarantees (under the Provably Approximately Correct framework) is one of the new frontier to reach - exemplified by the work of 1. Closed-loops systems with neural networks controllers are covered, as multiple case studies were presented at SAIV with controllers.