Yael Moros-Daval

Yael Moros-Daval

Auto-generated profile, based on publicly available information

Yael Moros-Daval

Valencian Research Institute for Artificial Intelligence (VRAIN), Universitat Politècnica de València, Valencia, Spain

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 September 2024 in Nature
Verified
Journal Article
Open access
PDF file
LZLexin Zhou
Lexin Zhou
WSWout Schellaert
Wout Schellaert
FMFernando Martínez-Plumed
Fernando Martínez-Plumed
YMYael Moros-Daval
Yael Moros-Daval
CFCèsar Ferri
Cèsar Ferri
Jose Hernandez-OralloJose Hernandez-Orallo
Jose Hernandez-Orallo
The prevailing methods to make large language models more powerful and amenable have been based on continuous scaling up (that is, increasing their size, data volume and computational resources1) and bespoke shaping up (including post-filtering2,3, fine tuning or use of human feedback4,5). However, larger and more instructable large language models may have become less reliable. By studying the relationship between difficulty concordance, task avoidance and prompting stability of several language model families, here we show that easy instances for human participants are also easy for the models, but scaled-up, shaped-up models do not secure areas of low difficulty in which either the model does not err or human supervision can spot the errors. We also find that early models often avoid user questions but scaled-up, shaped-up models tend to give an apparently sensible yet wrong answer much more often, including errors on difficult questions that human supervisors frequently overlook. Moreover, we observe that stability to different natural phrasings of the same question is improved by scaling-up and shaping-up interventions, but pockets of variability persist across difficulty levels. These findings highlight the need for a fundamental shift in the design and development of general-purpose artificial intelligence, particularly in high-stakes areas for which a predictable distribution of errors is paramount.

1 April 2026 in Nature
Verified
Journal Article
Open access
PDF file
LZLexin Zhou
Lexin Zhou
LPLorenzo Pacchiardi
Lorenzo Pacchiardi
FMFernando Martínez-Plumed
Fernando Martínez-Plumed
KCKatherine M. Collins
Katherine M. Collins
YMYael Moros-Daval
Yael Moros-Daval
Jose Hernandez-OralloJose Hernandez-Orallo
Jose Hernandez-Orallo
Ensuring safe and effective use of artificial intelligence (AI) requires understanding and anticipating its performance on new tasks, from advanced scientific challenges to transformed workplace activities 1–3 . So far, benchmarking has guided progress in AI but has offered limited explanatory and predictive power for general-purpose AI systems 4–8 , attributed to limited transferability across specific tasks 9–11 . Here we introduce general scales for AI evaluation that elicit demand profiles explaining what capabilities common AI benchmarks truly measure, extract ability profiles quantifying the general strengths and limits of AI systems and robustly predict AI performance for new task instances. Our fully automated methodology builds on 18 rubrics, capturing a broad range of cognitive and intellectual demands, which place different task instances on the same general scales, illustrated on 15 large language models (LLMs) and 63 tasks. Both the demand and the ability profiles on these scales bring new insights such as construct validity through benchmark sensitivity and specificity and explain conflicting claims about whether AI has reasoning capabilities. Ultimately, high predictive power at the instance level becomes possible using the general scales, providing superior estimates over strong black-box baseline predictors, especially in out-of-distribution settings (new tasks and benchmarks). The scales, rubrics, battery, techniques and results presented here constitute a solid foundation for a science of AI evaluation, underpinning the reliable deployment of AI in the years ahead.

30 January 2026 in Artificial Intelligence
Verified
Journal Article
Open access
LZLexin Zhou
Lexin Zhou
PCPablo A.M. Casares
Pablo A.M. Casares
FMFernando Martínez-Plumed
Fernando Martínez-Plumed
JBJohn Burden
John Burden
RBRyan Burnell
Ryan Burnell
Jose Hernandez-OralloJose Hernandez-Orallo
Jose Hernandez-Orallo

17 July 2025
Verified
Conference Paper
Closed access
DADaniel Romero Alvarado
Daniel Romero Alvarado
YDYael Moros Daval
Yael Moros Daval
BMBehzad Mehrbakhsh
Behzad Mehrbakhsh
FMFernando Martínez-Plumed
Fernando Martínez-Plumed
Este artículo describe el diseño, implementación y éxito del curso introductorio al análisis de datos y aprendizaje automático desarrollado por la unidad de formación de personal (UFASU) de la Universitat Politècnica de València desde 2024. Dirigido a personal universitario de todos los perfiles y titulaciones, el curso utiliza la plataforma RapidMiner para introducir técnicas fundamentales de ciencia de datos y aprendizaje automático sin necesidad de programación. Se analiza la metodología docente empleada, casos de estudio interdisciplinarios, resultados de participación, encuestas de satisfacción, desempeño en actividades y proyectos, y se presentan conclusiones y recomendaciones para replicar esta experiencia educativa innovadora.

17 July 2025
Verified
Conference Paper
Closed access
YDYael Moros Daval
Yael Moros Daval
DADaniel Romero Alvarado
Daniel Romero Alvarado
BMBehzad Mehrbakhsh
Behzad Mehrbakhsh
FMFernando Martínez-Plumed
Fernando Martínez-Plumed
Este artículo analiza el profundo impacto de la IA en la educación superior, destacando tanto sus enormes beneficios como los retos éticos y prácticos que implica. Partiendo del auge de herramientas generativas como ChatGPT, se examina mediante un análisis de la literatura cómo la IA puede personalizar el aprendizaje, automatizar tareas administrativas y docentes, e impulsar metodologías innovadoras en investigación. Al mismo tiempo, se abordan inquietudes sobre integridad académica, plagio, sesgos, privacidad de datos y fiabilidad de los sistemas inteligentes. Finalmente, se proponen recomendaciones para una integración responsable que incluya políticas claras, formación en alfabetización digital y rediseño de estrategias pedagógicas, configurando así a la IA como una aliada clave para potenciar el rol del profesorado y el éxito estudiantil.