AutoViewMem Moves Memory Disentanglement From Retrieval Time to Write Time

arXiv 2609.21940·medium signalXBlueskyLinkedInCopy link

Long-term conversational memory systems mix preferences, events, constraints and temporal updates into one representation, and the resulting semantic interference makes top-K retrieval noise-sensitive. AutoViewMem discovers candidate semantic views from interaction traces, selects a compact low-overlap view set, and uses those views to guide write-time structured extraction of provenance-grounded memories, plus offline consolidation for compactness. Because the disentangling happens at write time, plain top-K similarity search retrieves focused evidence with no routing or iterative retrieval at inference; it beats strong memory baselines on LoCoMo and PersonaMem under Qwen3-8B and Qwen3-14B.