Thrilled to announce, that our paper with Christos Diou, “Gradient-Guided Annealing for Domain Generalization”, has been accepted in CVRP2025! In our research, we tackle the problem of Domain Generalization (DG) from a gradient perspective, observing that conflicting gradients in datasets with diverse samples cause models to converge to suboptimal parameter configurations.The proposed Gradient-Guided Annealing (GGA) algorithm identifies loss surface minima that exhibit improved robustness by iteratively annealing its parameters, searching for points where gradients align across domains. You can learn more in the early access paper on ArXiv. Code available on GitHub.