Handover of In-Context Learning State Across Session Boundaries

研究会话边界处的交接问题：当上下文达到模型输入上限应用重启或由另一个 agent 接手任务时，应用必须决定把先前会话的哪些信息传下去作者把交接形式化为任务相对的 ICL 状态迁移，区分对早前内容的精确恢复与对目标分布的保持；在外生性条件下证明预测等价刻画了最粗的确定性充分交接，并给出固定长度比特需求提出三段式记录：决策与约束精确保存，重复证据用任务相关的统计量表示，效果无法被统计量保留的原始观测则原样保留高斯线性回归给出精确的有限维交接与有限比特扰动界，非参数回归则给出记忆量与平方预测误差关系的上下界

Handover of In-Context Learning State Across Session Boundaries

ID: d4289937
原文链接: https://arxiv.org/abs/2608.14528
PDF: https://arxiv.org/pdf/2608.14528v1
作者: Masahiro Kato, Taka Kato
日期: 2026-08-14
更新: 2026-08-14
分类: agents
来源类型: paper
标签: agent-memory, context-engineering, session-handover, in-context-learning, arxiv
质量评分: 4/5
抓取时间: 2026-08-18T05:28:55Z

中文导读

为什么值得关注

把会话交接形式化为 ICL 状态迁移，证明预测等价刻画最粗充分交接并给出固定比特记忆需求

该论文发表于 2026-08-14，作者为 Masahiro Kato, Taka Kato，arXiv 分类 cs.AI, econ.EM, math.ST, stat.ME, stat.ML；以上判断基于论文摘要所述内容。

关键信息

论文标题： Handover of In-Context Learning State Across Session Boundaries
作者： Masahiro Kato, Taka Kato
arXiv： https://arxiv.org/abs/2608.14528
发布时间： 2026-08-14
arXiv 分类： cs.AI, econ.EM, math.ST, stat.ME, stat.ML
备注： N/A
关联标签： agent-memory, context-engineering, session-handover, in-context-learning, arxiv

English Abstract

This study investigates the methodological and theoretical properties of session handover in applications that use large language models. A task may continue in a new session when the context reaches the model's input limit, when the application restarts, or when another agent is asked to finish the task. The application must then decide which information from the earlier session to pass on. We formulate handover as the transfer of a task-relative in-context learning (ICL) state and distinguish exact recovery of earlier material from preservation of the target distribution. Under an exogeneity condition, predictive equivalence characterizes the coarsest deterministic sufficient handover and gives a fixed-length bit requirement. The analysis isolates the effects of the memory constraint, the writer, and the continuation procedure, and quantifies the cost of writing before the realized downstream query is known. We propose a three-part record that stores decisions and constraints exactly, uses task-justified statistics for repeated evidence, and retains original observations whose effect is not preserved by those statistics. Gaussian linear regression gives an exact finite-dimensional handover and finite-bit perturbation bounds, while nonparametric regression gives upper and lower bounds that relate memory to squared prediction error. These results provide a theory and method for deciding what a handover must retain and how its memory requirement depends on the continuation task.

English Summary

This study investigates the methodological and theoretical properties of session handover in LLM applications: when context reaches the input limit, the application restarts, or another agent finishes the task, the application must decide which information from the earlier session to pass on. The authors formulate handover as the transfer of a task-relative in-context learning (ICL) state, distinguish exact recovery from preservation of the target distribution, and show that under an exogeneity condition predictive equivalence characterizes the coarsest deterministic sufficient handover with a fixed-length bit requirement. They propose a three-part record storing decisions and constraints exactly, task-justified statistics for repeated evidence, and original observations whose effect is not preserved by those statistics....

Obsidian Notes

内容由 opencli arxiv paper 拉取 arXiv 元数据与摘要生成。
中文导读与价值判断均锚定在条目已有摘要、论文摘要、作者、日期与分类信息上；未补充论文摘要之外的实验细节。