APDMem: Agent-Controlled Progressive Disclosure for Query-Adaptive Long-Term Memory

APDMem：用于查询自适应长期记忆的智能体控制渐进式披露机制

Abstract: Personalized LLM assistants must recover sparse evidence from long conversation histories across queries of varying complexity. We introduce APDMem (Agent-controlled Progressive Disclosure Memory), a hierarchical long-term memory architecture that applies progressive disclosure to memory retrieval.

摘要： 个性化大语言模型（LLM）助手必须能够从长对话历史中，针对不同复杂程度的查询提取稀疏的证据。我们引入了 APDMem（智能体控制渐进式披露记忆），这是一种将渐进式披露应用于记忆检索的分层长期记忆架构。

Rather than relying on a flat memory store or fixed retrieval granularity, APDMem represents conversation history as four progressively detailed layers: thematic summaries, personalized key facts, turn-level evidence notes, and raw messages.

APDMem 不再依赖扁平化的记忆存储或固定的检索粒度，而是将对话历史表示为四个渐进式详细层级：主题摘要、个性化关键事实、轮次级证据笔记以及原始消息。

At inference time, a controller applies progressive disclosure to the memory hierarchy: it first reads high-level summaries and drills into finer evidence only when needed. This creates an adaptive cost-fidelity trade-off: simple queries can terminate early, while complex temporal, multi-hop, or exact-evidence queries trigger deeper inspection.

在推理阶段，控制器会对记忆层级应用渐进式披露：它首先读取高层级摘要，仅在需要时才深入挖掘更细致的证据。这创造了一种自适应的成本与保真度权衡机制：简单查询可以提前终止，而复杂的时序查询、多跳查询或精确证据查询则会触发更深层次的检查。

A note synthesizer converts retrieved evidence into a query-focused structure that consolidates facts, orders events, and flags contradictions before final answer generation. Experiments on LongMemEval show that APDMem achieves strong performance for long-context memory reasoning while accessing only 8% of the total conversations.

笔记合成器会将检索到的证据转换为以查询为中心的结构，在最终生成答案之前整合事实、梳理事件顺序并标记矛盾点。在 LongMemEval 上的实验表明，APDMem 在长上下文记忆推理方面表现出色，且仅需访问总对话量的 8%。