arXiv：cs.AI· Yang Li, Hai Liu, Dian Shao, Yu Wang, Xiyu Chen, Di Wan, Sergey Volkov, Bozhi Wang, Ziyu Sun, Sihang Liu, Ye Luo, Xiaowei Zhang·· 21 小时前
Agent-UCT：用树状置信上界优化成本感知的智能体工作流
Agent-UCT: Upper Confidence Bounds Applied to Trees for Cost-Aware Agentic Workflow Optimization
AI 导读
AgentUCT 将剩余评估成本的经济模型引入 UCT 选择、并以原始质量做回传，在有限固定奖励树与持久缓存条件下证明了 T 次完整评估累积质量遗憾的 O(log T) 上界。
正文
Authors:Yang Li, Hai Liu, Dian Shao, Yu Wang, Xiyu Chen, Di Wan, Sergey Volkov, Bozhi Wang, Ziyu Sun, Sihang Liu, Ye Luo, Xiaowei Zhang
Abstract:Workflow search pays for complete quality observations, whose execution cost changes as reusable prefixes accumulate. AgentUCT uses an economic model of the remaining evaluation cost to guide UCT selection while backing up raw quality. Query-prefix accounting connects the search decisions to reusable execution state. For finite fixed-reward trees with bounded adaptive cold-cost estimates and successful persistent caching, we prove an $O(\log T)$ bound on cumulative quality regret over $T$ complete evaluations. Across five paired seeds, AgentUCT achieves 69.57% test F1 versus 68.01% for UCT at a 32M weighted-token budget on a retrospective HotpotQA panel. On GSM8K, it reaches 93.07% mean accuracy at the first budget, compared with 92.60% for UCT and 91.60% for AFlow. Single-factor RAG controls show that online prices and historical continuations improve the selected workflows at equal retained evaluation counts.
| Comments: | 69 pages, including supplementary material |
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.24162 [cs.AI] |
| (or arXiv:2607.24162v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.24162 arXiv-issued DOI via DataCite |
Submission history
From: Yang Li [view email]
[v1]
Mon, 27 Jul 2026 08:41:18 UTC (16 KB)
[v2]
Tue, 28 Jul 2026 18:08:04 UTC (16 KB)
[v3]
Sat, 10 Oct 2026 06:40:46 UTC (669 KB)
来源：arXiv：cs.AI · arxiv.org