\Delta$ Robustness

Δ-robustness focuses on ensuring the reliability and stability of algorithms and explanations under model variations or perturbations. Current research emphasizes developing methods to guarantee robustness, particularly in machine learning contexts, using techniques like interval abstractions and modifications to existing algorithms (e.g., bisection search) to achieve provable guarantees. This work is significant for improving the trustworthiness and reliability of AI systems and optimization algorithms, addressing concerns about their sensitivity to changes in input data or model parameters. The development of robust algorithms has implications for various applications, including explainable AI and robotics.