CGM-Rec Framework for Adaptive Recommendation Under Intent Drift

CGM-Rec introduces a continual graph memory framework to handle intent drift in recommendation systems. The architecture utilizes a conservative Semantic Graph Memory for stable knowledge and a fast-reacting Episodic Lesson Memory to learn from recent user feedback and failures.