Saved in:
Bibliographic Details
Main Authors: Zhang, Hanlong, Yang, Jingsheng, Li, Hao, He, Yuhao, Gong, Franck
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.08877
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913938849923072
author Zhang, Hanlong
Yang, Jingsheng
Li, Hao
He, Yuhao
Gong, Franck
author_facet Zhang, Hanlong
Yang, Jingsheng
Li, Hao
He, Yuhao
Gong, Franck
contents Function Calling is a crucial technique that enables Large Language Models (LLMs) to interact with external systems through APIs. However, the high latency associated with LLM-based Function Calling significantly impacts user experience. This paper presents a novel approach called Oriented Distillation for Inline Acceleration (ODIA) that leverages online user interaction data to accelerate Function Calling. By automatically identifying "simple queries" from production traffic and distilling knowledge from larger models to smaller ones, our method reduces response latency by 45% (expected) and 78% (median) while maintaining accuracy. We demonstrate the effectiveness of our approach through real-world deployment in a music application, where the smaller model successfully handles 60% of traffic with negligible accuracy loss. Our method requires minimal human intervention and continuously improves through automated data collection and model updating, making it a practical solution for production environments.
format Preprint
id arxiv_https___arxiv_org_abs_2507_08877
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ODIA: Oriented Distillation for Inline Acceleration of LLM-based Function Calling
Zhang, Hanlong
Yang, Jingsheng
Li, Hao
He, Yuhao
Gong, Franck
Machine Learning
Artificial Intelligence
Function Calling is a crucial technique that enables Large Language Models (LLMs) to interact with external systems through APIs. However, the high latency associated with LLM-based Function Calling significantly impacts user experience. This paper presents a novel approach called Oriented Distillation for Inline Acceleration (ODIA) that leverages online user interaction data to accelerate Function Calling. By automatically identifying "simple queries" from production traffic and distilling knowledge from larger models to smaller ones, our method reduces response latency by 45% (expected) and 78% (median) while maintaining accuracy. We demonstrate the effectiveness of our approach through real-world deployment in a music application, where the smaller model successfully handles 60% of traffic with negligible accuracy loss. Our method requires minimal human intervention and continuously improves through automated data collection and model updating, making it a practical solution for production environments.
title ODIA: Oriented Distillation for Inline Acceleration of LLM-based Function Calling
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2507.08877