Research

Papers

These projects represent our commitment to advancing the theoretical and practical foundations of AI to create more capable, intuitive, and user-aligned systems. Explore our research to learn more!

• 

  Fine-Tuning Next-Scale Visual Autoregressive Models with Group Relative Policy Optimization

  Authors: Matteo Gallici, Haitz Sáez de Ocáriz Borde
  Fine-tuning next-scale visual autoregressive models with reinforcement learning enhances image quality, style control, and generalization beyond pre-trained distributions.
• 

  Beyond Parallelism: Synergistic Computational Graph Effects in Multi-Head Attention

  Authors: Haitz Sáez de Ocáriz Borde
  Multi-head attention can induce synergistic interactions within the computational graph that transcend the efficiency gains attributable solely to parallelization.
• 

  Re-LAION–Caption 19M for Improved Text Alignment Fine-tuning

  Authors: Nicholas Merchant*, Haitz Sáez de Ocáriz Borde*, Andrei Cristian Popescu, Carlos Garcia Jurado Suarez
  Generative models struggle with prompt adherence due to noisy training data, but refining datasets with structured captions improves text-image alignment and reduces reliance on prompt engineering.