arXiv cs.AI· Tamara Stojanovska, Elisabeth Kirsten, Leon Swazinna, Muhammad Bilal Zafar·· 1 天前AI 评分58
Lost in Delegation：生成式搜索中任务编排的有效性研究
Lost in Delegation: Studying the Effectiveness of Task Orchestration in Generative Search
AI 导读
arXiv 论文《Lost in Delegation》研究生成式搜索中的任务编排效果，作者用按知识截止时间前后切分的数据集对比多种委派模式，发现委派相比非委派设置没有稳定的准确率提升，效果随模型和任务变化。当任务涉及编排 LLM 自身无法掌握的知识时，委派会明显降低准确率，而同一 LLM 直接使用工具表现更好。作者将委派查询失真、工具使用低效、智能体之间未能保留和使用信息列为反复出现的失败模式。
正文 · 原文
Abstract:LLMs have quickly evolved from models used in isolation to agentic systems that delegate tasks to external tools and other LLMs to solve complex assignments. This agentic augmentation has led to impressive capability gains. However, when and how delegation improves performance remains poorly understood. For example, it is unclear whether delegation helps when tasks involve post-cutoff information that the orchestrating LLM cannot reliably resolve from its own knowledge. In this work, we use purpose-built datasets with pre- and post-cutoff splits to compare various delegation modes. We find that delegation provides no consistent accuracy improvement over non-delegated setups, with effects varying across models and tasks. When tasks involve knowledge unavailable to the orchestrator, delegation can substantially degrade accuracy, while direct tool use with the same LLM performs much better. Our analysis identifies distorted delegated queries, inefficient tool use, and failures to preserve and use information between agents as recurring failure modes. These findings suggest that effective orchestration requires both well-calibrated delegation decisions and effective exchange of information between agents.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.16526 [cs.AI] |
| (or arXiv:2611.16526v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.16526 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Elisabeth Kirsten [view email]
[v1]
Thu, 12 Nov 2026 09:25:33 UTC (142 KB)
来源：arXiv cs.AI · arxiv.org