用关键词书签实现长对话记忆协作翻页，提升大模型长期对话质量。
Cooperative Memory Paging with Keyword Bookmarks for Long-Horizon LLM Conversations
- 用8-24词的关键词书签替代被丢弃的对话内容，支持按需召回。
- 在LoCoMo基准上四项指标领先，最高较基线提升8.7分（E2E）。
- 适合需要长期记忆的对话系统研发者，尤其关注记忆检索效率。
当大模型对话超出上下文窗口时，旧内容必须被替换——但如何在需要时恢复？本文提出协作翻页机制：将被驱逐的内容段落替换为极简关键词书签（[pN:keywords]，每条约8-24词），并赋予模型recall()工具按需检索完整内容。在包含10个真实多会话、300+轮次的LoCoMo基准上，该方法在六种对比方案中表现最优，超越截断、BM25、词重叠检索、搜索工具基线和全上下文策略，在四种模型（GPT-4o-mini, DeepSeek-v3.2, Claude Haiku, GLM-5）上均取得显著优势（四个独立LLM裁判评估，p=0.017，配对自助法）。通过5×4的消融实验（3,176个合成探针，1,600个LoCoMo探针），发现：(1) 固定大小粗粒度分页（fixed_20）达到96.7%性能，而基于主题变化的内容感知分页降至56.7%；(2) 淘汰策略依赖数据分布（合成数据用FIFO，LoCoMo用LFU）；(3) 两种书签生成策略优于启发式基线（分别提升+4.4和+8.7 E2E分）；(4) 当前瓶颈在于书签区分度不足——模型触发recall()率达96%，但仅57%情况下选对页面；关键词特异性单独导致25个百分点的准确率差异。
原文摘要 · Abstract (English)
When LLM conversations grow beyond the context window, old content must be evicted -- but how does the model recover it when needed? We propose cooperative paging: evicted segments are replaced with minimal keyword bookmarks ([pN:keywords], ~8-24 tokens each), and the model is given a recall() tool to retrieve full content on demand. On the LoCoMo benchmark (10 real multi-session conversations, 300+ turns), cooperative paging achieves the highest answer quality among six methods -- outperforming truncation, BM25, word-overlap retrieval, a search-tool baseline, and full context -- on four models (GPT-4o-mini, DeepSeek-v3.2, Claude Haiku, GLM-5), confirmed by four independent LLM judges ($p=0.017$, paired bootstrap). We then study the paging design space with a 5x4 ablation over boundary strategies and eviction policies (3,176 synthetic probes, 1,600 LoCoMo probes). Key findings: (1) coarse fixed-size pages (fixed_20) reach 96.7% while content-aware topic_shift collapses to 56.7%; (2) eviction policy choice is data-dependent (FIFO best on synthetic, LFU on LoCoMo); (3) two bookmark generation strategies improve over the heuristic baseline (+4.4 and +8.7 E2E points); (4) the remaining bottleneck is bookmark discrimination -- the model triggers recall() 96% of the time but selects the correct page only 57% when bookmarks are insufficiently distinctive. Keyword specificity alone accounts for a 25 percentage point accuracy difference.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。