Paper Number

ICIS2025-1718

Paper Type

Complete

Abstract

Supplier qualification assessment is pivotal for robust supply chain management, particularly in industries with complex production ecosystems. Traditional multi-dimensional indicator systems often suffer from subjective biases and fail to delineate specific strengths or weaknesses in supplier evaluations. This study proposes a novel framework integrating large language models (LLMs) with Chain-of-Thought (CoT) and Retrieval-Augmented Generation (RAG) technologies—termed Eval-RCoT—to enhance the precision and interpretability of supplier assessments. Leveraging a novel dataset from a major Chinese manufacturing enterprise, we evaluate the performance of ChatGPT and GLM in executing supplier evaluations. Our experiments reveal that Eval-RCoT effectively generates accurate rating levels and interpretable management insights. Quantitative validation via classification metrics and LLM-based metrics underscores the framework's reliability. These findings highlight the potential of LLM-driven approaches to mitigate subjectivity and streamline supplier evaluation processes in high-stakes industries, offering a scalable solution for dynamic supply chain demands.

Comments

07-DataAnalytics

Share

COinS
 
Dec 14th, 12:00 AM

Can large language models evaluate suppliers’ qualifications of complex production lines? An approach incorporating retrieval-augmented generation and Chain-of-Thought

Supplier qualification assessment is pivotal for robust supply chain management, particularly in industries with complex production ecosystems. Traditional multi-dimensional indicator systems often suffer from subjective biases and fail to delineate specific strengths or weaknesses in supplier evaluations. This study proposes a novel framework integrating large language models (LLMs) with Chain-of-Thought (CoT) and Retrieval-Augmented Generation (RAG) technologies—termed Eval-RCoT—to enhance the precision and interpretability of supplier assessments. Leveraging a novel dataset from a major Chinese manufacturing enterprise, we evaluate the performance of ChatGPT and GLM in executing supplier evaluations. Our experiments reveal that Eval-RCoT effectively generates accurate rating levels and interpretable management insights. Quantitative validation via classification metrics and LLM-based metrics underscores the framework's reliability. These findings highlight the potential of LLM-driven approaches to mitigate subjectivity and streamline supplier evaluation processes in high-stakes industries, offering a scalable solution for dynamic supply chain demands.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.