HyFunc : Accelerating LLM-based Function Calls for Agentic AI through Hybrid-Model Cascade and Dynamic Templating Conference: Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.1; August 09–13, 2026; Jeju Island, Republic of Korea Proceedings of the 32nd ACM SIGKD

Weibin Liao, Jian-guang Lou, Haoyi Xiong · 2026

Read HyFunc : Accelerating LLM-based Function Calls for Agentic AI through Hybrid-Model Cascade and Dynamic Templating Conference: Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.1; August 09–13, 2026; Jeju Island, Republic of Korea Proceedings of the 32nd ACM SIGKD by Weibin Liao, Jian-guang Lou, Haoyi Xiong (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-13665.txt.

Description

Agentic AI turns user intent into structured function calls, but three redundancies inflate latency. They are reprocessing a large function library per request, using a large slow model for whole predictable token sequences, and regenerating fixed boilerplate parameter syntax. HyFunc eliminates all three with a hybrid-model cascade. A large model distills intent into a single soft token, which guides a lightweight retriever to pick relevant functions and steers a smaller prefix-tuned model to generate the final call.

Dynamic templating injects boilerplate syntax on the fly inside an extended vLLM engine, removing syntactic redundancy. On the unseen BFCL benchmark, HyFunc reached 0.828 seconds latency, beating all baselines, with 80.1 percent performance, surpassing models of comparable scale. Weibin Liao leads the work with two co-authors.

Catalog record

CREATORWeibin Liao, Jian-guang Lou, Haoyi Xiong
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE55 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-13665

Browse the full library · About the Tower of Babel