Abstract:Function calling agents powered by Large Language Models (LLMs) select external tools to automate complex tasks. On-device agents typically use a retrieval module to select relevant tools, improving performance and reducing context length. However, existing retrieval methods rely on static and limited inputs, failing to capture multi-step tool dependencies and evolving task context. This limitation often introduces irrelevant tools that mislead the agent, degrading efficiency and accuracy. We propose Dynamic Tool Dependency Retrieval (DTDR), a lightweight retrieval method that conditions on both the initial query and the evolving tool calling plan. DTDR models tool dependencies from function calling demonstrations, enabling adaptive retrieval as plans unfold. We benchmark DTDR against state-of-the-art retrieval methods across multiple datasets and LLM backbones, evaluating retrieval precision, downstream task accuracy, and computational efficiency. Additionally, we explore strategies to integrate retrieved tools into prompts. Our results show that DTDR improves function calling success rates between $23\%$ and $104\%$ compared to state-of-the-art static retrievers.
| Comments: | 24 pages, 6 figures, 8 tables |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2512.17052 [cs.LG] |
| (or arXiv:2512.17052v4 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2512.17052 arXiv-issued DOI via DataCite |
From: Bhrij Patel [view email]
[v1]
Thu, 18 Dec 2025 20:40:25 UTC (687 KB)
[v2]
Tue, 23 Dec 2025 04:20:29 UTC (687 KB)
[v3]
Sun, 1 Mar 2026 13:46:47 UTC (690 KB)
[v4]
Fri, 17 Apr 2026 13:55:44 UTC (1,004 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。