TY  - RPRT
TI  - When Both Layers Learn: Training Dynamics of Representing Linear Models via ReLU Networks
AU  - Berk Tinaz
AU  - Changzhi Xie
AU  - Mahdi Soltanolkotabi
PY  - 2026
UR  - https://arxiv.org/abs/2606.04476
ID  - 2606.04476
ER  -