Evaluating LLM-Based Resume Information Extraction: A Comparative Study of Zero-Shot and One-Shot Learning Approaches in Portuguese-Specific and Multi-Language LLMs

Arthur Soares de Quadros, Arthur Soares de Quadros, Wesley Nogueira Galvão, Wesley Nogueira Galvão, Victória Emanuela Alves Oliveira, Victória Emanuela Alves Oliveira, Alessandro Vieira, Wladmir Cardoso Brandão, Wladmir Cardoso Brandão

2025

Abstract

This paper presents a comprehensive evaluation of Large Language Models (LLMs) in the task of information extraction from unstructured resumes in Portuguese. We examine six models, including both multilingual and Portuguese-specific variants, using 0-shot and 1-shot prompting strategies. To assess accuracy, we employ two complementary metrics: cosine similarity between model predictions and ground truth, and a composite LLM-as-a-Judge metric that weights factual information, semantic information, and order of components. Additionally, we analyze token cost and execution time to assess the practicality of each solution in production environments. Our results indicate that Gemini 2.5 Pro consistently achieves the highest accuracy, particularly under 1-shot prompting. GPT 4.1 Mini and GPT 4o Mini provide strong cost-performance trade-offs. Portuguese-specific models like Sabiá 3 achieves high average accuracy specially on 0-shot considering the cosine similarity metric. We also demonstrate how the inclusion of sections frequently missing in real resumes can significantly distort model evaluation. Our findings help determine model selection strategies for real-world applications involving semi-structured document parsing in a context of resume information extraction.

Paper Citation

in Harvard Style

Soares de Quadros A., Galvão W., Oliveira V., Vieira A. and Brandão W. (2025). Evaluating LLM-Based Resume Information Extraction: A Comparative Study of Zero-Shot and One-Shot Learning Approaches in Portuguese-Specific and Multi-Language LLMs. In Proceedings of the 21st International Conference on Web Information Systems and Technologies - Volume 1: WEBIST; ISBN 978-989-758-772-6, SciTePress, pages 313-324. DOI: 10.5220/0013837900003985

in Bibtex Style

@conference{webist25,
author={Arthur Soares de Quadros and Wesley Galvão and Victória Oliveira and Alessandro Vieira and Wladmir Brandão},
title={Evaluating LLM-Based Resume Information Extraction: A Comparative Study of Zero-Shot and One-Shot Learning Approaches in Portuguese-Specific and Multi-Language LLMs},
booktitle={Proceedings of the 21st International Conference on Web Information Systems and Technologies - Volume 1: WEBIST},
year={2025},
pages={313-324},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013837900003985},
isbn={978-989-758-772-6},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 21st International Conference on Web Information Systems and Technologies - Volume 1: WEBIST
TI - Evaluating LLM-Based Resume Information Extraction: A Comparative Study of Zero-Shot and One-Shot Learning Approaches in Portuguese-Specific and Multi-Language LLMs
SN - 978-989-758-772-6
AU - Soares de Quadros A.
AU - Galvão W.
AU - Oliveira V.
AU - Vieira A.
AU - Brandão W.
PY - 2025
SP - 313
EP - 324
DO - 10.5220/0013837900003985
PB - SciTePress