An Empirical Study on Large Language Models in Accuracy and Robustness under Chinese Industrial Scenarios

TL;DR

This study tackles the problem of assessing accuracy and robustness of Chinese-facing large language models in industrial settings. It introduces a first-of-its-kind benchmark of 1,200 industry-specific questions across eight sectors and a metamorphic testing framework that generates 13,631 robustness variants across four stability categories and eight abilities. By evaluating 12 LLMs from nine vendors (eight local, four global), the work finds overall accuracy below $0.6$, with global models typically outperforming local ones in reasoning and open-ended tasks, while local models better handle industry terminology. The findings offer practical guidance for industrial deployment, highlighting sector- and ability-specific robustness gaps and suggesting directions for model development, data collection, and tooling to support manufacturing applications.

Abstract

Recent years have witnessed the rapid development of large language models (LLMs) in various domains. To better serve the large number of Chinese users, many commercial vendors in China have adopted localization strategies, training and providing local LLMs specifically customized for Chinese users. Furthermore, looking ahead, one of the key future applications of LLMs will be practical deployment in industrial production by enterprises and users in those sectors. However, the accuracy and robustness of LLMs in industrial scenarios have not been well studied. In this paper, we present a comprehensive empirical study on the accuracy and robustness of LLMs in the context of the Chinese industrial production area. We manually collected 1,200 domain-specific problems from 8 different industrial sectors to evaluate LLM accuracy. Furthermore, we designed a metamorphic testing framework containing four industrial-specific stability categories with eight abilities, totaling 13,631 questions with variants to evaluate LLM robustness. In total, we evaluated 9 different LLMs developed by Chinese vendors, as well as four different LLMs developed by global vendors. Our major findings include: (1) Current LLMs exhibit low accuracy in Chinese industrial contexts, with all LLMs scoring less than 0.6. (2) The robustness scores vary across industrial sectors, and local LLMs overall perform worse than global ones. (3) LLM robustness differs significantly across abilities. Global LLMs are more robust under logical-related variants, while advanced local LLMs perform better on problems related to understanding Chinese industrial terminology. Our study results provide valuable guidance for understanding and promoting the industrial domain capabilities of LLMs from both development and industrial enterprise perspectives. The results further motivate possible research directions and tooling support.

Figures (4)

• Figure 1: The accuracy of LLMs in the Chinese industrial scenarios.
• Figure 2: The example of the self-knowledge blind spot discovery of GPT-4. The correct answer is B, marked in red.
• Figure 3: The robustness scores for different LLMs on various industrial sectors.
• Figure 4: The accuracy of LLMs in the Chinese industrial scenarios.