The field of formal verification and optimization is moving towards more efficient and effective methods for ensuring the quality and safety of complex systems, including neural networks and cyber-physical systems. Researchers are exploring innovative approaches, such as adaptive branch-and-bound tree exploration and hierarchical safety abstract interpretation, to improve the performance and granularity of formal verification techniques. Additionally, there is a growing interest in leveraging meta-learning and dimensionality reduction to enhance black-box optimization and inverse problems in various domains. Noteworthy papers in this area include:

• Adaptive Branch-and-Bound Tree Exploration for Neural Network Verification, which proposes a novel verification approach that demonstrates significant speedups over state-of-the-art verifiers.
• Advancing Neural Network Verification through Hierarchical Safety Abstract Interpretation, which introduces a new problem formulation that enables assessing multiple safety levels during the formal verification process.