arXiv：cs.AI· Albert Sadowski, Jaros{\l}aw A. Chudziak·· 5 小时前AI 评分37
写入时解读：面向多目标智能体记忆的策略消融研究
Interpreting at Write Time: A Policy Ablation for Multi-Goal Agent Memory
AI 导读
针对长期运行助手需在多个常驻目标下压缩历史的问题，研究对比了三种记忆写入策略：无目标摘要、单一全目标摘要、以及每目标各写一份摘要并联合读取。在多个模型与事件流上固定读取步骤后，按目标分别写入的摘要在意相关性、完整性和准确性上胜出，而全目标摘要即使预算更高也不如无目标摘要。
正文
Abstract:A long-running assistant cannot keep everything it has seen, so it summarises. Summarising is not neutral: what is kept is chosen against some notion of what the record is for, and that choice is made once, before anyone knows which of the user's standing goals will ask. Goals rarely disagree about what happened. They disagree about which parts of it were worth the space. Once the history is too long to re-read, the summary replaces the stream, and whatever it left out is gone. We ask what a memory should summarise for when it serves several standing goals at once. Three policies answer differently: summarise with no goal in view, write one summary covering every goal, or write one summary per goal and read them together. We compare them across several models and event streams, holding the read step fixed so that only the write differs. The goals do pull apart: summaries written for different goals overlap each other less than a summary overlaps a rewrite of itself. Per-goal summaries win on relevance, completeness and accuracy, and the all-goal summary loses even to the neutral one written at a fraction of its budget. Interpreting at write pays off, but only for the goal that later asks.
| Comments: | Accepted to PALM workshop at NeurIPS 2026 |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.02897 [cs.AI] |
| (or arXiv:2610.02897v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.02897 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Albert Sadowski [view email]
[v1]
Fri, 2 Oct 2026 06:44:39 UTC (31 KB)
来源：arXiv：cs.AI · arxiv.org