Cold-Start Ranking: Architectural Semantic Labels Predict Memory Use Where Graph Methods Go Blind

Abstract

In agent memory systems, graph-based ranking methods go blind at cold-start: 91% of new nodes remain invisible to PageRank and access-count scores throughout the observation window. Architectural semantic labels assigned at node creation predict future organic use at AUC 0.84, a 34-point advantage over random, of which approximately 18 points exceed pure-text baselines (0.66). The most trivial baseline, newest-first, performs below random (AUC 0.41). With accumulated history, structural reinforcement takes over (AUC 0.917). We contribute: a quantified cold-start blind spot, a demonstration that architectural labels serve as cold-start priors, a disentanglement of label vs. text contribution, and a stated boundary. No cross-system superiority is claimed. Includes both English and Chinese versions.