MambaAct: Efficient Mamba Action Experts for Flow-Matching Vision-Language-Action Models

Key points

• Vision-language-action (VLA) models have shown strong promise for generalist robotic manipulation, but their inference cost remains a major bottleneck for real-time control.1
• This limitation is especially pronounced in flow-matching VLAs such as $π{0.5}$, where the action expert is evaluated repeatedly during iterative denoising.1
• MambaAct retains the pretrained visual-language prefix encoder of $π{0.5}$, while replacing its transformer-based action expert with a compact Mamba-based action expert.1
• To preserve visual-language conditioning without repeated token-level cross-attention, MambaAct uses a lightweight Context Compressor to map Vision-language model (VLM) prefix tokens into a compact conditioning signal that gates and modulates each Mamba action block.1

Sources (1)

• [1]MambaAct: Efficient Mamba Action Experts for Flow-Matching Vision-Language-Action Models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 16, 05:20 AM

  We introduce MambaAct, an efficient Mamba action expert for flow-matching VLA policies.

  Vision-language-action (VLA) models have shown strong promise for generalist robotic manipulation, but their inference cost remains a major bottleneck for real-time control.

Extractive summary: sentences quoted from the sources.