A Model-Agnostic Approach to Mitigate Gradient Interference for Multi-Task Learning | IEEE Journals & Magazine | IEEE Xplore