IACM-RL：面向动态意图波动下复杂工具调用的意图感知上下文管理与强化学习框架
IACM-RL: Intent-Aware Context Management and Reinforcement Learning for Complex Tool Invocation under Dynamic Intent Fluctuations
针对动态用户意图噪声导致长程工具调用出现意图偏离和重复 API 循环的问题，研究者提出 IACM-RL 框架，包含合成多场景轨迹的 DynamicIntent 流水线与五维诊断指标，并部署基于 BeliefState 的自生成上下文管理器，用结构化 stale flags 隔离被覆盖参数。
Authors:Dingwei Zhu, Jiahan Li, Chengjun Pan, Yunxian Yang, Yunbin Zhao, Yunke Zhang, Zhonghang Lu, Zhuohui Sheng, Chenhao Huang, Jiahang Lin, Yajie Yang, Junlin Shang, Shichun Liu, Yuhui Wang, Honglin Guo, Junjie Ye, Xin Guo, Jiazheng Zhang, Ming Zhang, Shihan Dou, Zhiheng Xi, Tao Gui, Qi Zhang, Xipeng Qiu, Xuanjing Huang
Abstract:Executing long-horizon tool invocations in real-world environments is severely challenged by dynamic user intent noise. Existing methods attempt robustness via implicit history scanning or text compression, yet predominantly assume perfect instructions in simplistic scenarios. Inevitably, under fluctuating contexts, obsolete constraints dilute model attention, triggering intent deviation and repetitive API loops. To resolve this, we propose IACM-RL, a comprehensive framework for robust tool invocation. First, we introduce the DynamicIntent pipeline, synthesizing trajectories across several fine-grained fluctuation scenarios, paired with a five-dimensional diagnostic metric suite. Second, IACM-RL deploys a BeliefState-based Self-Generated Context Manager that proactively tracks shifting goals and isolates overwritten parameters using structural stale flags. To autonomously internalize this state-tracking capability, we optimize the policy using a hierarchical intent-driven reward alongside two auxiliary losses (CM extraction and state distillation). Experiments on DynamicIntent, BFCL-V3, and $\mathrm{\tau}^2$-Bench demonstrate that IACM-RL outperforms baselines, reducing infinite loops and stale context errors while enhancing out-of-domain generalization.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.02110 [cs.CL] |
| (or arXiv:2608.02110v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2608.02110 arXiv-issued DOI via DataCite |
Submission history
From: Dingwei Zhu [view email]
[v1]
Mon, 3 Aug 2026 12:09:50 UTC (5,606 KB)
[v2]
Wed, 14 Oct 2026 11:48:21 UTC (5,387 KB)
来源：arXiv：cs.CL · arxiv.org