发现并检测大模型应用中隐藏的资源消耗循环，提升系统效率。
Unsupervised Cycle Detection in Agentic Applications
- 结合时间调用栈与语义相似性分析，无监督识别执行循环。
- 在1575条轨迹上达到F1 0.72，显著优于单一方法。
- 适合关注大模型应用性能优化的研究者与开发者。
由大语言模型驱动的智能体应用表现出非确定性行为，可能形成隐蔽的执行循环，无声消耗资源且不触发显式错误。传统可观测性平台无法检测此类高成本低效问题。我们提出一种无监督循环检测框架，融合结构与语义分析：首先通过计算高效的时序调用栈分析识别显式循环，再利用语义相似性分析发现由冗余内容生成引起的隐性循环。在基于LangGraph的股票市场应用中，1575条执行轨迹上的评估显示，该混合方法取得F1值0.72（精确率0.62，召回率0.86），显著优于单独使用结构方法（F1: 0.08）和语义方法（F1: 0.28）。尽管结果令人鼓舞，仍存在较大改进空间，未来工作需进一步优化该方法并解决其现有局限。
原文摘要 · Abstract (English)
Agentic applications powered by Large Language Models exhibit non-deterministic behaviors that can form hidden execution cycles, silently consuming resources without triggering explicit errors. Traditional observability platforms fail to detect these costly inefficiencies. We present an unsupervised cycle detection framework that combines structural and semantic analysis. Our approach first applies computationally efficient temporal call stack analysis to identify explicit loops and then leverages semantic similarity analysis to uncover subtle cycles characterized by redundant content generation. Evaluated on 1575 trajectories from a LangGraph-based stock market application, our hybrid approach achieves an F1 score of 0.72 (precision: 0.62, recall: 0.86), significantly outperforming individual structural (F1: 0.08) and semantic methods (F1: 0.28). While these results are encouraging, there remains substantial scope for improvement, and future work is needed to refine the approach and address its current limitations.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。