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

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

Browse the full ICLR paper archive.