Research

My research focuses on building AI systems that are trustworthy, fair, and transparent — with a particular emphasis on large language models and responsible AI.

Research Areas

AI Ethics & Trustworthiness

Developing frameworks, metrics, and tools to evaluate and improve the ethical behavior of AI systems. This includes defining Responsible AI benchmarks and building open-source toolkits for bias, fairness, and explainability assessment.

Large Language Models (LLMs)

Investigating the behavioral properties of LLMs including personality traits, hallucination patterns, and emoji generation. Developing end-to-end tools for hallucination mitigation and evaluation, and benchmarks for efficacy, robustness, and privacy.

Explainability & Interpretability

Creating model-agnostic metrics to evaluate the explainability of machine learning predictions. Focused on making black-box models more interpretable for stakeholders across industries.

Bias Mitigation & Fairness

Uncovering and addressing biases in AI systems — from face generation models to NLP pipelines. Integrating advanced bias mitigation techniques into production AI systems and open-source libraries.

Natural Language Processing

Building transformer-based models for relation extraction, sentiment analysis in Portuguese, and information extraction from corporate documents. Applied NLP to real-world problems including labor market analysis for the ILO.

Applied ML & Deep Learning

Scaling transformer-based models across 100+ GPUs, COVID-19 X-ray classification, voice pathology detection, remote sensing data classification, and financial pricing models using neural networks.

Publications

LibVulnWatch: A Deep Assessment Agent System and Leaderboard for Uncovering Hidden Vulnerabilities in Open-Source AI Libraries

Z. Wu, S. Cho, U. Mohammed, C. Munoz, K. Costa, X. Guan, T. King, Z. Wang, E. Kazim, A. Koshiyama

2025 ACL 2025 Student Research Workshop ICML 2025 TAIG Workshop arXiv:2505.08842

From Text to Emoji: How PEFT-Driven Personality Manipulation Unleashes the Emoji Potential in LLMs

N. Jain, Z. Wu, C. E. M. Villalobos, A. Hilliard, X. Guan, A. Koshiyama, E. Kazim, P. Treleaven

2025 Findings of NAACL 2025 NeurIPS 2024 Workshop on Behavioral ML

THaMES: An End-to-End Tool for Hallucination Mitigation and Evaluation in Large Language Models

M. Liang, A. Arun, Z. Wu, C. E. M. Villalobos, J. Lutch, E. Kazim, A. Koshiyama, P. Treleaven

2024 Workshop on Socially Responsible Language Modelling Research

Eliciting Big Five Personality Traits in Large Language Models: A Textual Analysis with Classifier-Driven Approach

A. Hilliard, C. E. M. Villalobos, Z. Wu, A. S. Koshiyama

2024 arXiv:2402.08341

Evaluating Explainability for Machine Learning Predictions Using Model-Agnostic Metrics

C. E. M. Villalobos, K. da Costa, B. Modenesi, A. S. Koshiyama

2024 arXiv:2302.12094

Sentimental Analysis on Social Media Comments with Recurring Models and Pretrained Word Embeddings in Brazilian Portuguese

C. E. M. Villalobos, L. Forero, H. De Mello, C. Valencia, A. Orjuela

2023 DOI: 10.1145/3582768

Cognitive Search: A Free Information Retrieval Web Service to Coronavirus Scientific Papers

J. D. Bermudez Castro, C. E. M. Villalobos, S. A. Canchumuni, L. Forero, M. A. C. Pacheco

2023 IEEE ColCACI DOI: 10.1109/ColCACI59285.2023.10225924

Uncovering Bias in Face Generation Models

C. E. M. Villalobos, S. Zannone, U. Mohammed, A. S. Koshiyama

2023 arXiv:2302.11562

A Free Web Service for Fast COVID-19 Classification of Chest X-Ray Images

J. D. Bermudez Castro, R. Rei, J. E. Ruiz, P. Achanccaray, S. A. Canchumuni, C. Munoz Villalobos et al.

2022 DOI: 10.1007/978-3-031-10522-7_29

Deep Learning for Mapping Rainwater Drainage Networks Using Remote Sensing Data

Potratz, Julia, C. E. M. Villalobos, S. W. A. Canchumuni, M. A. C. Pacheco

2021 DOI: 5774/4838

Improvement of Optical Character Recognition for Structured Documents using Generative Adversarial Networks

J. D. Bermudez Castro, S. W. A. Canchumuni, C. E. M. Villalobos et al.

2021 DOI: 10.1109/ICCSA54496.2021.00046

Petrolese: How to Build a Specialized Oil and Gas Corpus in Portuguese

F. Correa Cordeiro, C. E. M. Villalobos

2020 Rio Oil and Gas Expo and Conference

Construction of a Transformer-based Model for Relation Extraction

C. E. M. Villalobos, L. Mendoza, R. Tanscheit

2019 CBIC 2019

Conditional Pricing Model with Heteroscedasticity: Evaluation of Brazilian Funds

C. L. Santos DA, B. F. Fischberg Oliveira, F. L. Cyrino, C. E. M. Villalobos

2019 DOI: 10.1590/S0034-759020190402

Severe Asthma Exacerbations Prediction Using Neural Networks

A. Silveira, C. E. M. Villalobos, L. Mendoza

2019 DOI: 10.1007/978-3-030-20257-6_10

Analysis and Classification of Voice Pathologies using Glottal Signal Parameters with Recurrent Neural Networks and SVM

L. A. Forero Mendoza, C. E. M. Villalobos, M. Kohler, E. Batista, M. A. Pacheco

2019 DOI: 10.5220/0007250700190028

Improving Transfer Learning Performance: an Application in the Classification of Remote Sensing Data

G. L. Tenorio, C. E. M. Villalobos, L. A. Forero Mendoza, E. C. da Silva, W. Caarls

2019 DOI: 10.5220/0007372201740183