TY  - RPRT
TI  - Near-optimal and Efficient First-Order Algorithm for Multi-Task Learning with Shared Linear Representation
AU  - Shihong Ding
AU  - Fangyu Du
AU  - Cong Fang
PY  - 2026
UR  - https://arxiv.org/abs/2605.00473
ID  - 2605.00473
ER  -