A First Look at Bugs in LLM Inference Engines

TL;DR

This paper presents the first systematic empirical analysis of bugs in LLM inference engines, revealing common bug types, root causes, and providing insights for improving their reliability and development.

Contribution

It introduces a comprehensive dataset of 929 real-world bugs in LLM inference engines and offers a taxonomy of bug symptoms and root causes, advancing understanding in this area.

Findings

Six bug symptom types identified

28 root causes categorized

Guidelines for bug detection and fixing proposed

Abstract

Large language model-specific inference engines (in short as \emph{LLM inference engines}) have become a fundamental component of modern AI infrastructure, enabling the deployment of LLM-powered applications (LLM apps) across cloud and local devices. Despite their critical role, LLM inference engines are prone to bugs due to the immense resource demands of LLMs and the complexities of cross-platform compatibility. However, a systematic understanding of these bugs remains lacking. To bridge this gap, we present the first empirical study on bugs in LLM inference engines. We mine official repositories of 5 widely adopted LLM inference engines, constructing a comprehensive dataset of 929 real-world bugs. Through a rigorous open coding process, we analyze these bugs to uncover their symptoms, root causes, commonality, fix effort, fix strategies, and temporal evolution. Our findings reveal six bug symptom types and a taxonomy of 28 root causes, shedding light on the key challenges in bug detection and location within LLM inference engines. Based on these insights, we propose a series of actionable implications for researchers, inference engine vendors, and LLM app developers, along with general guidelines for developing LLM inference engines.