This paper has been withdrawn by Ziyang Liu
No PDF available, click to view other formats
Abstract: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.
| Comments: | The authors have decided to withdraw this version following internal review regarding authorship and contribution agreements |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2604.12376 [cs.CL] |
| (or arXiv:2604.12376v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2604.12376 arXiv-issued DOI via DataCite |
From: Ziyang Liu [view email]
[v1]
Tue, 14 Apr 2026 07:06:35 UTC (337 KB)
[v2]
Sun, 24 May 2026 03:41:37 UTC (1 KB) (withdrawn)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。