MambaAct: Efficient Mamba Action Experts for Flow-Matching Vision-Language-Action Models
This work introduces MambaAct, an efficient Mamba action expert for flow-matching VLA policies, which retains the pretrained visual-language prefix encoder of $\pi_{0.5}$, while replacing its transformer-based action expert with a compact Mamba-based action expert.
Khuyen Pham, Tran Nguyen Le, Edoardo Fazzari et al.
· 0 citations