TY  - RPRT
TI  - Backward Oversmoothing: why is it hard to train deep Graph Neural Networks?
AU  - Nicolas Keriven
PY  - 2026
UR  - https://arxiv.org/abs/2505.16736
ID  - 2505.16736
ER  -