arXiv：cs.AI· Yannan Wang, Longli Yang, Zhen Liu, Abhishek Kumar, Carsten Maple·· 19 小时前
CoMIC：面向 Web 云边系统长时程 LLM 智能体的协同记忆与洞察循环
CoMIC: Collaborative Memory and Insights Circulation for Long-Horizon LLM Agents in Web Cloud-Edge Systems
AI 导读
CoMIC 是一个无需更新参数的云边协同框架，采用“边缘执行、云端反思”范式，让边缘智能体以子目标导向的层级记忆在本地执行。云端 LLM 评审器异步评估经 Web 网关上传的轨迹，筛选可复用经验并以语义子目标标识索引，将跨智能体指导回传至所有边缘智能体。
正文
Abstract:LLM-based Web agents are increasingly deployed on edge servers to carry out long-horizon tasks on behalf of users with low latency. However, the lightweight models that fit on edge servers often struggle with long-horizon tasks that require persistent memory, subgoal tracking, and reflection. Fine-tuning edge models after deployment is costly and difficult to scale across heterogeneous nodes, while purely local execution confines each agent to its own isolated experience. We propose \textsc{CoMIC}, a parameter-update-free cloud-edge framework for Collaborative Memory and Insights Circulation. Following the paradigm of ``Act at the Edge, Reflect in the Cloud'', edge agents execute locally with subgoal-oriented hierarchical memory and selective re-expansion of relevant histories, while a cloud-side LLM critic asynchronously evaluates the trajectories uploaded through a Web gateway, admits reusable experience, and circulates cross-agent guidance indexed by semantic subgoal identifiers back to all edge agents. On five long-horizon benchmarks, \textsc{CoMIC} enables weak edge agents to achieve non-zero success on tasks where a standard memory-based agent fails entirely, and improves their task progress over both the standard agent and purely local execution while reducing context consumption relative to the standard agent, all without updating model parameters. When stronger heterogeneous agents join the edge, the weak agents further improve in success, progress, and grounding with lower context consumption.
| Comments: | This version updates the overview and framework figures, refines the experimental setup, provides a more in-depth analysis of the experimental results, and improves overall readability |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.00756 [cs.AI] |
| (or arXiv:2606.00756v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2606.00756 arXiv-issued DOI via DataCite |
Submission history
From: Yannan Wang [view email]
[v1]
Sat, 30 May 2026 14:45:39 UTC (2,014 KB)
[v2]
Tue, 13 Oct 2026 11:48:59 UTC (13,437 KB)
来源：arXiv：cs.AI · arxiv.org