arXiv cs.AI· Zhiyuan Huang, Shengan Zheng, Chenghu Yang, Song Zhang, Lingyun Li, Chen Yu, Weiquan Huang, Linpeng Huang·· 3 小时前AI 评分42
Synapse-Control：用轻量级运行时控制让智能体经验复用可执行
Synapse-Control: Making Agent Experience Actionable with Lightweight Runtime Control
AI 导读
论文提出 Synapse-Control，一种智能体运行时控制方案，用规则初筛建议、1.7B 微调分类器评分替代大模型生成式诊断。在 AppWorld 和 WebShop 每方法 1,308 次尝试中，它取得 270 次成功（比 Base 高 29.8%），替换 4,987 次生成式诊断调用，辅助控制时间较 Sentry 分别降低 50.2% 和 51.0%。
正文 · 原文
Abstract:Large language model agents reuse execution experience to improve reliability on complex tool-based tasks. Systems such as Sentry call large models to generate failure diagnoses before retrieving advice, adding inference cost to experience reuse. We present Synapse-Control, an agent runtime whose lightweight control plane selects advice and determines when to send it to the task-executing model. Rules shortlist relevant advice, and a lightweight 1.7B classifier fine-tuned on teacher-labeled state-suggestion pairs scores those candidates instead of generating a diagnosis. Micro-contracts encode reusable interface checks, such as required arguments or valid options, without prescribing complete task programs. A Prompt Value Table records earlier guidance and feedback to limit repeated suggestions through cooldown and token-budget checks. Across 1,308 attempts per method in AppWorld and WebShop, Synapse-Control achieves 270 successes (29.8% more than Base), replaces 4,987 generative diagnosis calls, and reduces auxiliary control time relative to Sentry by 50.2% and 51.0%, respectively.
| Comments: | 21 pages, 8 figures, 18 tables, including appendices |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2611.13615 [cs.CL] |
| (or arXiv:2611.13615v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.13615 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zhiyuan Huang [view email]
[v1]
Tue, 10 Nov 2026 05:50:12 UTC (189 KB)
来源：arXiv cs.AI · arxiv.org