π0-FAST
A vision-language-action (VLA) model that adopts an autoregressive action generation paradigm, discretizing actions into tokens and predicting them token by token, consistent with the way language models generate text.
Timeline (last 90 days)
Backtrack was evaluated on the LIBERO and MetaWorld benchmarks, covering three VLA models: SmolVLA, π0.5, and π0-FAST.
Flow-matching action experts (such as π0.5) model action generation as a continuous transformation process from noise to actions, producing smooth action sequences through iterative denoising and suiting high-dimensional continuous control; autoregressive methods (such as π0-FAST) discretize actions into tokens and predict them token by token.
All Facts (2)
Flow-matching action experts (such as π0.5) model action generation as a continuous transformation process from noise to actions, producing smooth action sequences through iterative denoising and suiting high-dimensional continuous control; autoregressive methods (such as π0-FAST) discretize actions into tokens and predict them token by token.
50%UnverifiedBacktrack was evaluated on the LIBERO and MetaWorld benchmarks, covering three VLA models: SmolVLA, π0.5, and π0-FAST.
50%