MoTVLA: A Vision-Language-Action Model with Unified Fast-Slow Reasoning

TL;DR

MoTVLA is a unified vision-language-action model that combines fast and slow reasoning to improve robot manipulation efficiency and language steerability, leveraging pre-trained models and domain-specific transformers.

Contribution

It introduces a mixture-of-transformers framework that integrates fast domain-specific reasoning with generalist perception, enhancing robot policy learning and language steerability.

Findings

Outperforms existing models in manipulation tasks

Improves language steerability in robotic policies

Demonstrates efficiency in real-world robot experiments

Abstract

Integrating visual-language instructions into visuomotor policies is gaining momentum in robot learning for enhancing open-world generalization. Despite promising advances, existing approaches face two challenges: limited language steerability when no generated reasoning is used as a condition, or significant inference latency when reasoning is incorporated. In this work, we introduce MoTVLA, a mixture-of-transformers (MoT)-based vision-language-action (VLA) model that integrates fast-slow unified reasoning with behavior policy learning. MoTVLA preserves the general intelligence of pre-trained VLMs (serving as the generalist) for tasks such as perception, scene understanding, and semantic planning, while incorporating a domain expert, a second transformer that shares knowledge with the pretrained VLM, to generate domain-specific fast reasoning (e.g., robot motion decomposition), thereby improving policy execution efficiency. By conditioning the action expert on decomposed motion instructions, MoTVLA can learn diverse behaviors and substantially improve language steerability. Extensive evaluations across natural language processing benchmarks, robotic simulation environments, and real-world experiments confirm the superiority of MoTVLA in both fast-slow reasoning and manipulation task performance.