Fresh Memory, Stale Plans: Dependency-Scoped Validation for Distributed LLM-Agent Memory

新鲜的记忆，陈旧的计划：分布式大模型智能体记忆的依赖范围验证

Abstract: Distributed LLM-agent teams can read the latest shared facts and still act on an obsolete plan. A planner may derive an action from requirement $r_3$, another agent may commit $r_4$, and an executor may receive $r_4$ without replacing the plan derived from $r_3$. We call this \emph{stale-plan execution}: state freshness does not establish that the plan authorizing an action remains valid.

摘要： 分布式大模型（LLM）智能体团队即使读取了最新的共享事实，仍可能基于过时的计划采取行动。例如，规划者可能根据需求 $r_3$ 推导出某项行动，而另一个智能体提交了 $r_4$，执行者在接收到 $r_4$ 后，却并未替换掉基于 $r_3$ 推导出的计划。我们将此称为“陈旧计划执行”（stale-plan execution）：状态的新鲜度并不能保证授权该行动的计划依然有效。

We introduce PlanFence, a dependency-scoped action-validation protocol. Plans cite the exact public records they used, and an executor validates only the records that can affect the pending external action, replanning once or blocking when validation is incomplete.

我们引入了 PlanFence，这是一种基于依赖范围的行动验证协议。计划会引用其所使用的确切公共记录，执行者仅验证那些可能影响待处理外部行动的记录；当验证不完整时，执行者会重新规划或进行阻塞。

In 30 controlled live workflows with a post-plan revision, a freshness-only executor acts on the obsolete plan in every task, whereas PlanFence completes all tasks without an invalid action. Controlled replay reveals two conditional boundaries: proactive synchronization yields lower coordination stall at low churn, while PlanFence avoids repeated update-path coordination as churn grows and avoids validating unrelated state as the shared keyspace grows. These are controlled safety and systems-cost results, not general task-accuracy gains.

在 30 个包含计划后修订的受控实时工作流中，仅依赖“状态新鲜度”的执行者在每项任务中都基于过时的计划采取了行动，而 PlanFence 在完成所有任务的同时未出现任何无效行动。受控重放揭示了两个条件边界：在低变动率（churn）下，主动同步能降低协调停顿；而随着变动率增加，PlanFence 避免了重复的更新路径协调，且随着共享键空间（keyspace）的增长，它避免了对无关状态的验证。这些是受控的安全性和系统成本结果，而非任务准确性的普遍提升。