MemReader: From Passive to Active Extraction for Long-Term Agent Memory

ArXi:2604.07877v1 Announce Type: new Long-term memory is fundamental for personalized and autonomous agents, yet populating it remains a bottleneck. Existing systems treat memory extraction as a one-shot, passive transcription from context to structured entries, which struggles with noisy dialogue, missing references, and cross-turn dependencies, leading to memory pollution, low-value writes, and inconsistency. In this paper, we