SimVLA: A Simple VLA Baseline for Robotic Manipulation

TL;DR

SimVLA introduces a minimal, transparent baseline for vision-language-action models in robotic manipulation, achieving state-of-the-art results with a simple design and standardized training, facilitating clearer attribution of improvements.

Contribution

It presents a streamlined, decoupled VLA baseline that outperforms larger models and standardizes evaluation, aiding future research clarity.

Findings

SimVLA outperforms multi-billion-parameter models on simulation benchmarks.

SimVLA achieves comparable real-robot performance to larger models.

A minimal design with 0.5B parameters suffices for strong manipulation performance.

Abstract

Vision-Language-Action (VLA) models have emerged as a promising paradigm for general-purpose robotic manipulation, leveraging large-scale pre-training to achieve strong performance. The field has rapidly evolved with additional spatial priors and diverse architectural innovations. However, these advancements are often accompanied by varying training recipes and implementation details, which can make it challenging to disentangle the precise source of empirical gains. In this work, we introduce SimVLA, a streamlined baseline designed to establish a transparent reference point for VLA research. By strictly decoupling perception from control, using a standard vision-language backbone and a lightweight action head, and standardizing critical training dynamics, we demonstrate that a minimal design can achieve state-of-the-art performance. Despite having only 0.5B parameters, SimVLA outperforms multi-billion-parameter models on standard simulation benchmarks without robot pretraining. SimVLA also reaches on-par real-robot performance compared to pi0.5. Our results establish SimVLA as a robust, reproducible baseline that enables clear attribution of empirical gains to future architectural innovations. Website: https://frontierrobo.github.io/SimVLA