InterFuse framework improves payment risk ranking through adaptive model fusion
November 3, 2026
InterFuse utilizes a shared fusion state to guide transaction-adaptive aggregation of heterogeneous models. By encoding model representations, prediction scores, and input coverage as tokens, the framework handles unavailable outputs and improves ranking accuracy for limited-budget transaction reviews.
HOW THIS AFFECTS YOU
●
builderYou can improve ensemble model performance in production by using availability-aware attention to weight model outputs dynamically.