发表机构
University of Washington; Paul G. Allen School of Computer Science & Engineering（华盛顿大学; 保罗·G·艾伦计算机科学与工程学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究提出与运行时无关的持久智能体架构，定义连续性不变量与相关协议，通过 Enoch 实现该设计，验证其可跨模型等变更保留智能体连续性。
AI 中文摘要
智能体系统通常由当前产生其行为的模型和编排器来描述。这种边界对于单次执行很有用，但对于长期存在的智能体而言，其可能在变更模型、编排智能体、交互会话和主机服务器的同时保留单一身份、记忆和可执行代码谱系，这就导致该描述存在规范不足的问题。我们提出一种与运行时无关的持久智能体架构。承载连续性的基底 $P_t=(I_t,M_t,B_t)$ 包含架构身份表示、私有持久记忆和版本化软件主体。可替换的部署绑定包含执行基底 $E_t=(R_t,H_t,D_t)$，其提供推理器、编排器和主机，以及一组交互界面 $S_t$，例如聊天、API 或用户界面绑定。已部署的执行体为 $A_t=P_t\triangleright(E_t,S_t)$；当授权协议在受控部署边界内保留可追溯谱系并转移延续权限时，变更任一可替换层均为迁移，而非智能体创建。我们定义六个连续性不变量及“暂停-检查点-验证-绑定-恢复-重启”协议。Enoch 将该设计实现为可复用主体，加上私有已安装身份、记忆、工作流状态和延续权限，基础设施依赖项置于版本化提供商合同之后。对冻结公共提交的洁净室运行通过了 833 项核心测试，以及与核心套件分开执行的 92 项提供商和库测试；部署已验证了推理器版本、交互界面和主机机器的替换，同时保留了承载连续性的状态。该证据支持机械可替换性和授权系统连续性，而非行为不变性或详尽的成对评估。下游测量问题在于，授权的延续体是否仍能回忆、组合和执行其身份。
英文摘要
Long-lived AI agents may replace their models, orchestration harnesses, interaction surfaces and hosts while retaining identity, memory and executable body lineage. We present a runtime-independent architecture that separates a continuity-bearing substrate $P_t=(I_t,M_t,B_t)$ from replaceable execution and interaction bindings. Six continuity invariants and a quiesce--checkpoint--validate--bind--rehydrate--resume protocol govern state preservation, capability changes and continuation authority. Enoch implements the design through a reusable software body, private installed state, provider contracts and fenced migration. We report complementary evidence from frozen implementation snapshots and three kinds of live study. An operational case retains an established instance's identity, nonempty memory, body revision and historical task IDs across host, joint Codex-to-Claude model/harness, and chat-surface substitutions. A separate cross-host task resumes from a verified artifact checkpoint. A supervised Codex--Muse--Codex study completes five ordinary round trips and five matched same-runtime controls on their first attempts after a separate qualification run. A planned worker interruption recovers through the native task API with the checkpoint preserved, a stale reply rejected and one task completion recorded. These results support mechanical continuity for the tested deployments and bounded workflows. They do not establish behavioral equivalence, arbitrary-task portability, general exactly-once external effects or unattended reliability.
Comments11 pages, 3 figures, 5 tables. Adds Claude/host/surface migrations, checkpointed tasks, supervised Muse round trips, controls, fault recovery, an invariant-evidence map and PAI-Bench reference. Clarifies authority and recovery scope. Public artifacts: https://github.com/our-ark/enoch-muse-runtime
