Large Auto Regressive Model

Large autoregressive models (LARMs) are a class of machine learning models that predict sequential data, such as text, images, or even street networks, one element at a time. Current research focuses on improving their efficiency (e.g., through data distillation techniques), expanding their capabilities to handle diverse data modalities (including integrating text and images), and mitigating their environmental impact by accurately estimating and reducing their carbon footprint. These advancements are driving progress in various fields, from embodied AI and sustainable AI development to the generation of realistic city-scale simulations.