Hybrid Robustness Verification for Spatio-Temporal Neural Networks

arXiv·low signal

As AI is deployed into safety-critical systems, formal robustness guarantees on model behavior become essential. This paper presents a hybrid verification method for spatio-temporal neural networks, combining techniques to certify robustness over time-series and video inputs. It is of interest to teams deploying perception or control agents in regulated, safety-critical settings where certification—not just empirical testing—is required.