MARM: Enhancing Recommendation Systems by Caching Computation Results for Multi-Layer Attention

MARM leverages caching to overcome computational limitations in recommendation systems, enabling multi-layer attention modeling of user history for improved accuracy without significant performance degradation.