#vision-language-action reasoning

NVIDIA's ThinkAct framework introduces a dual-system approach combining vision-language reasoning with reinforced visual latent planning, significantly improving robot manipulation and planning in complex environments.