KEAT-TemporalGNN

Repository for AAAI-2026 paper Kernelized Edge Attention: Addressing Semantic Attention Blurring in Temporal Graph Neural Networks

---

Create and activate the Environment

Refer to environment.yml for package details.

TGN-KEAT

Model Training

KEAT_TGN/train.py

Model Testing

KEAT_TGN/test.py

Results on tgbl-wiki are in:

KEAT_TGN/keat_tgn.ipynb

TGN-DyGFormer

We follow original DyGFormer steps for training and evaluation : KEAT_DyGFormer/README.md

Code Credits

KEAT codebase is built on top of following repos:

• TGB
• DyGFormer
• LeTE