Portrait de Adriana Romero Soriano

Adriana Romero Soriano

Membre industriel principal
Chaire en IA Canada-CIFAR
Professeure adjointe, McGill University
Chercheuse scientifique, Meta AI Research (FAIR)
Sujets de recherche
Apprentissage profond
Modèles génératifs
Vision par ordinateur

Biographie

Adriana Romero-Soriano est chercheuse à Meta (FAIR, Fundamental AI Research), professeure adjointe à l'Université McGill, membre industriel principal de Mila – Institut québécois d’intelligence artificielle et titulaire d'une chaire en IA Canada-CIFAR. Ses recherches se situent à l'intersection des modèles génératifs, de la vision par ordinateur et de l'IA responsable. Ses travaux les plus récents portent sur l'amélioration de la qualité, de la contrôlabilité, de la cohérence et de la diversité de représentation des systèmes de création de contenu visuel. Elle a obtenu son doctorat à l'Université de Barcelone, où elle a travaillé avec Carlo Gatta, et a été chercheuse postdoctorale pendant deux ans à Mila, où elle a travaillé avec le professeur Yoshua Bengio.

Étudiants actuels

Reyhane Askari Hemmat
Collaborateur·rice de recherche - UdeM
reyhane.askari.hemmat@mila.quebec
Sumana Basu
Collaborateur·rice alumni - McGill
Superviseur⋅e principal⋅e :
Doina Precup
sumana.basu@mail.mcgill.ca
Site web
Github
Google Scholar
Aarash Feizi
Doctorat - McGill
Superviseur⋅e principal⋅e :
Reihaneh Rabbany
feiziaar@mila.quebec
Site web
Github
Google Scholar
Soumya Sharma
Doctorat - McGill
Superviseur⋅e principal⋅e :
Golnoosh Farnadi
soumya.sharma@mila.quebec
Site web
Github
Google Scholar

Publications

RoboJEPA: Scaling Robotic Latent World Models
Artem Zholus
Nicolas Beltran-Velez
Jianhao Yuan
Sarath Chandar
Tushar Nagarajan
Daniel Severo
Koustuv Sinha
Michal Drozdzal
Adriana Romero Soriano
Jeannette Bohg
Nicolas Ballas
Mahmoud Assran
Latent world models have shown a remarkable ability to predict future states and to plan in the real world. In practice, however, we lack a … (voir plus)principled way to estimate how their capabilities scale with model size, data, and compute, an open problem that slows progress in the field. In this work we present RoboJEPA, a world model based on the Joint Embedding Predictive Architecture (JEPA) and trained on a large-scale dataset spanning 12 robotic embodiments. We show that RoboJEPA's imagination error, the error of its latent rollouts, follows a second-order power law in compute, allowing us to predict model quality well beyond the scale at which the law is fit. We further show that downstream robotic planning performance improves predictably with compute, and that imagination error is strongly correlated with it, making it a reliable proxy for real-robot evaluation. Finally, we demonstrate that latent world models can be deployed zero-shot as robotic agents, planning toward a single goal image to solve tasks requiring long-horizon planning on real hardware. We release all model checkpoints together with our training and robot deployment code. To our knowledge, this is the first work to establish scaling laws for multi-embodiment robotic world models trained on real robot data, and RoboJEPA, at 8B parameters, is the largest JEPA predictor model trained to date.
2026-10-06
arXiv (prépublication)
doi.org
arxiv.org
DIG In: Evaluating Disparities in Image Generations with Indicators for Geographic Diversity
Melissa Hall
Candace Ross
Adina Williams
Nicolas Carion
Michal Drozdzal
Adriana Romero Soriano
The unprecedented photorealistic results achieved by recent text-to-image generative systems and their increasing use as plug-and-play conte… (voir plus)nt creation solutions make it crucial to understand their potential biases. In this work, we introduce three indicators to evaluate the realism, diversity and prompt-generation consistency of text-to-image generative systems when prompted to generate objects from across the world. Our indicators complement qualitative analysis of the broader impact of such systems by enabling automatic and efficient benchmarking of geographic disparities, an important step towards building responsible visual content creation systems. We use our proposed indicators to analyze potential geographic biases in state-of-the-art visual content creation systems and find that: (1) models have less realism and diversity of generations when prompting for Africa and West Asia than Europe, (2) prompting with geographic information comes at a cost to prompt-consistency and diversity of generated images, and (3) models exhibit more region-level disparities for some objects than others. Perhaps most interestingly, our indicators suggest that progress in image generation quality has come at the cost of real-world geographic representation. Our comprehensive evaluation constitutes a crucial step towards ensuring a positive experience of visual content creation for everyone.
2024-05-06
International Conference on Learning Representations (Accept (Poster))
doi.org
openreview.net