Recon: Reducing Conflicting Gradients From the Root For Multi-Task Learning

Guangyuan Shi, Qimai Li, Wenlong Zhang, Jiaxin Chen, Xiao-Ming Wu