Full Professor, Institute of Software,Chinese Academy of Sciences
1 paper at NeurIPS 2025
Transformers fall short of surpassing the performance ceiling of GNNs, but introducing appropriate constraints can effectively enhance the generalization of GNNs.