Research
Research proposes LaLa-World, a language-conditioned latent world model for planning and robot policies
LaLa-World pairs a JEPA-style latent world model with a language model and trains across simulation, robot, image and video data. It reportedly matches a video generation baseline's planning performance at 6x lower training cost and 13.9x faster planning, and it performs competitively with Cosmos P…
Every outlet that ran it 1 article
Hugging Face PapersResearchLead
LaLa-World: Scaling Latent World Models with Natural Language Actions