The method ensemble learning with calibrated losses that we recently developed has been accepted for publication at ICML 2025. For further details, please refer to the arXiv and GitHub links.