AI RESEARCH
HTM-EAR: Importance-Preserving Tiered Memory with Hybrid Routing under Saturation
arXiv CS.AI
•
ArXi:2603.10032v1 Announce Type: cross Memory constraints in long-running agents require structured management of accumulated facts while preserving essential information under bounded context limits. We We evaluate the system under sustained saturation (15,000 facts; L1 capacity 500; L2 capacity 5000) using synthetic streams across five random seeds and real BGL system logs. Ablation studies compare the full system against variants without cross-encoder re-ranking, without routing gates, with LRU eviction, and an oracle with unbounded memory.