Public articles linked to the same research event.
arXiv The work presents Hermes, a dual-scale architecture in which Micro-GT applies relation-aware attention over a temporally safe heterogeneous session neighborhood and Macro-GT attends over non-anticipative ecosystem climate tokens plus adaptive class prototypes; on more than 130 million high-risk transaction sessions it achieves a 44.44% relative reduction in customer friction (FR@50R) and a 24.66% relative improvement in fraud recall (R@5FR) over the production XGBoost system, with 85.12% overall ROC-AUC.
The work presents Hermes, a dual-scale architecture in which Micro-GT applies relation-aware attention over a temporally safe heterogeneous session neighborhood and Macro-GT attends over non-anticipative ecosystem climate tokens plus adaptive class prototypes; on more than 130 million high-risk transaction sessions it achieves a 44.44% relative reduction in customer friction (FR@50R) and a 24.66% relative improvement in fraud recall (R@5FR) over the production XGBoost system, with 85.12% overall ROC-AUC.
The work presents Hermes, a dual-scale architecture in which Micro-GT applies relation-aware attention over a temporally safe heterogeneous session neighborhood and Macro-GT attends over non-anticipative ecosystem climate tokens plus adaptive class prototypes; on more than 130 million high-risk transaction sessions it achieves a 44.44% relative reduction in customer friction (FR@50R) and a 24.66% relative improvement in fraud recall (R@5FR) over the production XGBoost system, with 85.12% overall ROC-AUC.
The work presents Hermes, a dual-scale architecture in which Micro-GT applies relation-aware attention over a temporally safe heterogeneous session neighborhood and Macro-GT attends over non-anticipative ecosystem climate tokens plus adaptive class prototypes; on more than 130 million high-risk transaction sessions it achieves a 44.44% relative reduction in customer friction (FR@50R) and a 24.66% relative improvement in fraud recall (R@5FR) over the production XGBoost system, with 85.12% overall ROC-AUC.