arXiv：cs.LG· Jiahang Sun, Zhiwei Shang, Zhipiao Liu, Hongwei Yang, Guoqing Xie, Shuang Qiu, Zhongxiang Dai·· 13 小时前
CB-Orchestrator：用上下文赌博机为 LLM 智能体做自适应工作流优化
CB-Orchestrator: Adaptive Workflow Optimization for LLM Agents via Contextual Bandits
AI 导读
CB-Orchestrator 通过把工作流选择建模为 Contextual Bandit 问题，将工作流生成与选择解耦，从而减少训练阶段所需的反馈。在五个基准测试上它均优于所有基线，MATH 数据集上训练阶段 API token 开销降低不少于 40.55%，含工作流池生成在内的端到端训练 token 开销降低不少于 29.73%。
正文
Abstract:Large Language Model (LLM) agents have demonstrated remarkable capabilities in tackling complex tasks through agentic workflows. However, manually designing these workflows is labor-intensive and lacks flexibility to adapt to diverse queries. While automated workflow optimization is a promising direction, existing methods often incur expensive API costs during training-phase feedback collection. Furthermore, they fail to adaptively reuse high-quality workflows, which creates a bottleneck for both performance and cost-effectiveness. To address these limitations, we propose CB-Orchestrator, an adaptive framework that decouples workflow generation from selection. In the first stage, we construct a diverse pool of candidate workflows via evolutionary search. In the second stage, we formulate workflow selection as a Contextual Bandit problem, which enables sample-efficient learning by balancing exploration and exploitation, thereby significantly reducing the feedback required for training. During inference, the model adaptively selects the optimal workflow tailored to each query. Evaluations across five diverse benchmarks demonstrate that CB-Orchestrator consistently outperforms all baselines. Notably, on the MATH dataset, CB-Orchestrator achieves better performance than strong recent baselines while reducing training-phase API token overhead by no less than 40.55% and total end-to-end training (including workflow pool generation) token overhead by no less than 29.73%.
| Comments: | Presented at ICML 2026 Workshop on Compositional Learning: Safety, Interpretability, and Agents |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.13064 [cs.LG] |
| (or arXiv:2610.13064v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.13064 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zhongxiang Dai [view email]
[v1]
Fri, 9 Oct 2026 06:25:19 UTC (494 KB)
来源：arXiv：cs.LG · arxiv.org