Guido Ivetta
I am Guido Ivetta, a PhD candidate in Computer Science and Agencia I+D+i Doctoral Fellow in the Contextualized AI (C-AI) Research Group at the Facultad de Matemática, Astronomía, Física y Computación (FaMAF), Universidad Nacional de Córdoba (UNC) in Argentina, supervised by Luciana Benotti. I graduated as Abanderado (Flag Bearer / Valedictorian of FaMAF, GPA 9.78/10) with the 3rd highest overall GPA across the entire Universidad Nacional de Córdoba, and I serve as an Assistant Professor (Profesor Asistente) in Computer Science at FaMAF UNC.
In addition to my doctoral research at UNC, I am the Technical Lead of the Artificial Intelligence Team at Fundación Vía Libre, a Student Member of the ACL Ethics Committee (since September 2024), and was a visiting researcher at Imperial College London (January – April 2025) under the supervision of Rafael A. Calvo.
Research Interests
My research focuses on evaluating and improving the reliability, fairness, and cultural grounding of modern artificial intelligence systems:
- LLM Calibration & Selective Prediction: Developing methods to estimate model uncertainty and confidence in generative language and vision-language models, enabling systems to selectively abstain when uncertain. (Winner of the Best Student Paper Award at LatinX in AI @ NAACL 2024).
- Bias, Fairness & Stereotyping in AI: Designing auditing methodologies and tools like EDIA to systematically detect, characterize, and evaluate demographic biases and regional stereotypes across Latin American and under-represented communities.
- Participatory AI & Community Co-Design: Bridging technical evaluation and civil society by creating community-sourced benchmarks like HESEIA—co-designed with 370 educators and 5,000+ secondary students across 189 schools—and LACES.
- Multilingual & Multimodal Benchmarking: Assessing the cross-cultural and dialectal competence of vision-language models and machine translation across regional Spanish varieties and global languages (e.g., CVQA, La Leaderboard, CaMMT).
Recent Highlights
- EMNLP 2026 (Main Conference): Co-authored “INCLUDE-2.0: Disentangling Knowledge Gaps in Multilingual LLM Evaluation”, a large-scale region-aware factual benchmark covering 91 languages from 86 countries across 220K samples.
- ACL 2026 (San Diego): Presented “Adaptive Data Collection for Latin-American Community-sourced Evaluation of Stereotypes (LACES)” at Findings of ACL 2026, presented orally at the StereACuLT workshop, and co-organized the Birds-of-a-Feather (BoF) session “Who Gets Left Behind? Identity, Culture, and Power in NLP Systems” with Alberto Testoni (Amsterdam UMC).
- AI and Ethics (Springer Nature, 2026): Published “Towards culturally-appropriate AI for health in the majority world: insights from participatory work in Latin America”.
- ICLR 2026 (Rio de Janeiro): Our paper on “Logical Reasoning Evaluation and Social Bias” was selected as a featured work at the ICLR 2026 Logical Reasoning Workshop.
- India AI Impact Summit 2026 (New Delhi): Represented Fundación Vía Libre presenting HESEIA within the AI Research & Impact Symposium.
- EMNLP 2025 (Suzhou): Presented “HESEIA: A community-based dataset for evaluating social biases in large language models, co-designed in real school settings in Latin America” in the main conference oral program.
- ACL 2025 (Vienna): Co-instructed the tutorial “Navigating Ethical Challenges in NLP: Hands-on strategies for students and researchers” and co-authored “La Leaderboard”.
- Communications of the ACM (2024): Published “Exploring stereotypes and biases in language technologies in Latin America” in CACM.
- ACM FAccT 2024 (Rio de Janeiro): Conducted the hands-on tutorial and demonstration of EDIA (“Evaluation of Bias for Experts in Discrimination, not in Computer Science”).
Navigation & Resources
Curriculum Vitae English PDF Español PDF - Publications: Journal articles, conference papers, tutorials, and preprints.
- Talks & Presentations: Invited presentations, conference talks, tutorials, and workshops.
- Teaching: University appointments at FaMAF UNC and tutorial instruction.
- Portfolio & Projects: Open-source benchmarks, datasets, and collaborative initiatives.
