All terms

Glossary · AI & ML

Large behavior model

Also known as: LBM

In brief

A large behavior model is a foundation-scale neural network trained to predict actions from observations across many tasks and embodiments. The robotics analogue of a large language model: pretrain on a wide corpus of robot data, then fine-tune for specific tasks.

Where a large language model predicts the next token, a large behavior model (LBM) predicts the next action. The training data is robot demonstrations, teleoperation logs, and simulation rollouts across many tasks and platforms. The bet is that scale + diversity produces a base policy that generalizes, the same way GPT-4 generalizes from its training corpus.

LBMs are upstream of VLA models — once you have an LBM you can specialize it via fine-tuning, RLHF-equivalent techniques, or task-specific imitation. Toyota Research Institute, Physical Intelligence, and Skild AI are notable LBM-focused organizations as of 2026.

Related terms

See it in the wild

Browse robots and brands using these techniques

Glossary entries are upstream. The catalog is where the implementations live.