Sitemap
A list of all the posts and pages found on the site. For search engines and AI systems, an XML Sitemap, a concise LLM Context Summary (llms.txt), a comprehensive LLM Full Knowledge Base (llms-full.txt), and downloadable Curriculum Vitae PDFs in English and Spanish are available.
Pages
Posts
portfolio
CVQA: Culturally-Diverse Multilingual VQA
A globally diverse visual question answering benchmark covering 30+ languages to test vision-language models on cultural grounding.
EDIA & EDIA 2.0: Evaluation of Discrimination in AI
Open-source toolkit and platform for evaluating gender, racial, and regional stereotypes in language and multimodal technologies.
Ethics in AI Research at Fundación Vía Libre
Advancing algorithmic transparency, fairness, and public-interest AI research in Latin America.
HESEIA: Co-Designed Benchmark for Social Biases in LLMs
A community-based dataset of 46,499 sentences co-designed in real school settings across Latin America to evaluate social biases in LLMs.
INCLUDE-2.0: Region-Aware Multilingual LLM Benchmark
A large-scale benchmark for region-aware multilingual factual evaluation spanning 91 languages from 86 countries across 220K samples.
La Leaderboard: Spanish Varieties & Regional Languages LLM Benchmark
A comprehensive open leaderboard and evaluation suite for Spanish dialectal varieties and co-official regional languages.
LACES: Latin-American Community-Sourced Evaluation of Stereotypes
An adaptive data collection protocol to surface and benchmark regional stereotypes across Latin America.
publications
A Methodology to Characterize Bias and Harmful Stereotypes in Natural Language Processing in Latin America
Published in arXiv preprint arXiv:2207.06591, 2022
Presents a pioneering methodology to characterize harmful stereotypes and demographic biases across Spanish language technologies in Latin America.
Aprendizaje Automático Aplicado a la Huerta Familiar Agroecológica en la Patagonia
Published in Tesina de Licenciatura en Ciencias de la Computación, FaMAF, Universidad Nacional de Córdoba, 2023
Tesina de Licenciatura en Ciencias de la Computación sobre técnicas de machine learning aplicadas al manejo agroecológico de huertas familiares.
Inteligencia Artificial Aplicada a la Huerta Familiar Agroecológica en la Patagonia
Published in Revista de Ciencias Sociales, Universidad Nacional de Quilmes (UNQ), 2023
Artículo publicado en la Revista de Ciencias Sociales (UNQ) sobre la aplicación situada de herramientas de inteligencia artificial para huertas agroecológicas en la Patagonia.
Your Stereotypical Mileage May Vary: Practical Challenges of Evaluating Biases in Multiple Languages and Cultural Contexts
Published in Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), 2024
Examines practical challenges of evaluating social biases in NLP across multiple languages and socio-cultural settings, illustrating how regional stereotypes cannot simply be translated from English benchmarks.
CVQA: Culturally-diverse Multilingual Visual Question Answering Benchmark
Published in Proceedings of the 38th International Conference on Neural Information Processing Systems (NeurIPS 2024), 2024
Visual Question Answering (VQA) is a crucial task in multimodal AI, often used to test vision-language models on understanding visual and linguistic information. CVQA introduces a culturally-diverse multilingual benchmark covering 30+ languages to evaluate model generalization across diverse cultural contexts.
Detecting Correct Answers to Open Questions and Its Impact on Language Models’ Confidence Scores
Published in LatinX in AI (LXAI) Research Workshop at NAACL 2024, Mexico City, 2024
Explores how the automatic detection of correct open-ended answers influences model confidence scoring and calibration in Large Language Models.
Exploring Stereotypes and Biases in Language Technologies in Latin America
Published in Communications of the ACM, 67(8), 54-56, 2024
Discusses the impacts of automated language technologies in Latin America and highlights participatory methodologies to uncover and mitigate local biases.
Selectively Answering Visual Questions
Published in Findings of the Association for Computational Linguistics: ACL 2024, 2024
Investigates selective prediction in multimodal models, enabling large vision-language models to abstain from answering questions when their confidence is low.
CaMMT: Benchmarking Culturally Aware Multimodal Machine Translation
Published in Findings of the Association for Computational Linguistics: EMNLP 2025, 2025
Introduces CaMMT, an evaluation benchmark for culturally aware multimodal machine translation spanning diverse cultural concepts, images, and languages.
ROSA: Addressing Text Understanding Challenges in Photographs via ROtated SAmpling
Published in arXiv preprint arXiv:2506.03665, 2025
Introduces ROtated SAmpling (ROSA) to significantly improve vision-language model robustness when reading rotated or misoriented text in photographs taken by visually impaired users.
La Leaderboard: A Large Language Model Leaderboard for Spanish Varieties and Languages of Spain and Latin America
Published in Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), 2025
An open evaluation benchmark and leaderboard tracking language model performance across regional varieties of Spanish and co-official languages of Spain and Latin America.
Navigating Ethical Challenges in NLP: Hands-on Strategies for Students and Researchers
Published in Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025 Tutorial Abstracts), 2025
A tutorial presented at ACL 2025 providing hands-on strategies and ethical frameworks for NLP practitioners, students, and reviewers.
Synthetic Social Data: Trials and Tribulations
Published in arXiv preprint arXiv:2510.19952, 2025
Critically examines the opportunities and ethical pitfalls of using Large Language Models to generate synthetic conversational data for social and behavioral science.
HESEIA: A Community-Based Dataset for Evaluating Social Biases in Large Language Models, Co-Designed in Real School Settings in Latin America
Published in Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP 2025), 2025
Presents HESEIA, a benchmark for evaluating social biases in LLMs created through participatory co-design workshops with secondary school students in Latin America.
Insights from a Disaggregated Analysis of Kinds of Biases in a Multicultural Dataset
Published in Proceedings of the 9th Widening NLP Workshop (WiNLP 2025), 2025
Presents a fine-grained disaggregated evaluation of specific stereotypes and biases within multicultural datasets, revealing disparate impacts across demographic axes.
Sesgos en Modelos de Lenguaje y Construcción Participativa de Benchmarks de Evaluación
Published in Trabajo Final de Licenciatura, FaMAF, Universidad Nacional de Córdoba, 2025
Trabajo Final de Grado de Licenciatura en Ciencias de la Computación de Sofía Martinelli, dirigido por Guido Ivetta.
Logical Reasoning Evaluation and Social Bias
Published in ICLR 2026 Workshop on Logical Reasoning of Large Language Models, 2026
Reviews how stereotypes and socio-cultural priors interact with reasoning chains and formal deduction in Large Language Models.
Adaptive Data Collection for Latin-American Community-sourced Evaluation of Stereotypes (LACES)
Published in Findings of the Association for Computational Linguistics: ACL 2026, 2026
Proposes LACES, an adaptive community-sourced protocol for collecting and evaluating regional stereotypes in Latin America.
Towards Culturally-Appropriate AI for Health in the Majority World: Insights from Participatory Work in Latin America
Published in AI and Ethics, Springer, 6(4), 345, 2026
Investigates requirements and ethical considerations for conversational health AI in Latin America through participatory research with citizens and clinicians.
INCLUDE-2.0: Disentangling Knowledge Gaps in Multilingual LLM Evaluation
Published in Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026), 2026
A large-scale benchmark for region-aware multilingual factual evaluation covering 91 languages from 86 countries with over 220K curated samples.
talks
Evaluating Bias in Language Models with the EDIA Tool
Published:
Tool demonstration and workshop on identifying algorithmic bias with EDIA.
Workshop on Responsible AI & Bias Evaluation
Published:
Main organizer and speaker; awarded KhipuX / Americas Fund grantee funding.
EDIA: Evaluation of Bias for Experts in Discrimination, Not in Computer Science
Published:
Interactive tutorial and CRAFT session on auditing social biases in NLP using the open-source EDIA tool.
Detecting Correct Answers to Open Questions and Its Impact on Language Models’ Confidence Scores
Published:
Presentation on calibration, uncertainty estimation, and evaluation of open question answering in LLMs.
Socio-Cultural Evaluation of AI
Published:
Hands-on tutorial and practical session on socio-cultural evaluation of stereotypes and biases in language models.
Navigating Ethical Challenges in NLP: Hands-on Strategies for Students and Researchers
Published:
Tutorial at ACL 2025 covering ethical issues in NLP, hands-on strategies for evaluation, and responsible research practices.
HESEIA: A Community-Based Dataset for Evaluating Social Biases in Large Language Models
Published:
Main conference oral and poster presentation at EMNLP 2025 in Suzhou, China.
Insights from a Disaggregated Analysis of Kinds of Biases in a Multicultural Dataset
Published:
Presentation on fine-grained disaggregated evaluation of specific stereotypes in multicultural datasets.
HESEIA at the India AI Impact Summit 2026
Published:
Presented HESEIA and participatory community co-design methodologies for evaluating social biases in LLMs.
Logical Reasoning Evaluation and Social Bias
Published:
Oral and poster presentation selected as a featured work at the ICLR 2026 Logical Reasoning Workshop.
Logical Reasoning Evaluation and Social Bias
Published:
Talk examining the interaction between formal logical reasoning in LLMs and socio-cultural bias traces.
Adaptive Data Collection for Latin-American Community-sourced Evaluation of Stereotypes (LACES)
Published:
Findings poster, oral presentation at StereACuLT workshop, and co-organizer of BoF ‘Who Gets Left Behind? Identity, Culture, and Power in NLP Systems’.
Practical Workshop on Socio-Cultural AI Evaluation
Published:
Hands-on workshop leader instructing on evaluating socio-cultural biases, regional stereotypes, and calibration in language models.
teaching
Profesor Asistente (Assistant Professor) in Computer Science
Undergraduate Courses, FaMAF, Universidad Nacional de Córdoba, 2021
Assistant Professor (Profesor Asistente) in the Computer Science Section at the Facultad de Matemática, Astronomía, Física y Computación (FaMAF), Universidad Nacional de Córdoba (March 2021 – Present).
Thesis Evaluation Jury (Jurado de Tesis de Grado)
Thesis Examination Committee (Jurado Evaluador), FaMAF, Universidad Nacional de Córdoba, 2025
Served as a Thesis Committee Member / Examiner (Jurado de Tesis de Grado) for Licenciatura en Ciencias de la Computación theses at FaMAF, Universidad Nacional de Córdoba (2025 onwards):
Instructor & Tutorial Co-organizer: Ethics and Socio-Cultural AI Evaluation
Tutorials & Academic Summer Schools, ACL 2025 & Khipu, 2025
Co-instructor and tutorial organizer delivering hands-on instruction for international researchers, graduate students, and practitioners:
Thesis Advisor: Lic. Sofía Martinelli
Undergraduate Thesis Supervision (Dirección de Tesis), FaMAF, Universidad Nacional de Córdoba, 2025
Thesis Advisor (Director de Trabajo Final de Grado) for the Licenciatura en Ciencias de la Computación degree at FaMAF, Universidad Nacional de Córdoba:
