Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success

TL;DR

This paper presents an optimized fine-tuning approach for vision-language-action models that significantly improves their efficiency, success rates, and flexibility in robotic tasks, demonstrated through state-of-the-art results in simulation and real-world experiments.

Contribution

The paper introduces a comprehensive fine-tuning recipe for VLAs, including parallel decoding, action chunking, continuous action representation, and L1 regression, leading to OpenVLA-OFT with superior performance.

Findings

Achieved 97.1% success rate on LIBERO benchmark, up from 76.5%.

Increased action generation throughput by 26 times.

Enabled high-frequency dexterous control on a bimanual robot.

Abstract

Recent vision-language-action models (VLAs) build upon pretrained vision-language models and leverage diverse robot datasets to demonstrate strong task execution, language following ability, and semantic generalization. Despite these successes, VLAs struggle with novel robot setups and require fine-tuning to achieve good performance, yet how to most effectively fine-tune them is unclear given many possible strategies. In this work, we study key VLA adaptation design choices such as different action decoding schemes, action representations, and learning objectives for fine-tuning, using OpenVLA as our representative base model. Our empirical analysis informs an Optimized Fine-Tuning (OFT) recipe that integrates parallel decoding, action chunking, a continuous action representation, and a simple L1 regression-based learning objective to altogether improve inference efficiency, policy performance, and flexibility in the model's input-output specifications. We propose OpenVLA-OFT, an instantiation of this recipe, which sets a new state of the art on the LIBERO simulation benchmark, significantly boosting OpenVLA's average success rate across four task suites from 76.5% to 97.1% while increasing action generation throughput by 26$\times$. In real-world evaluations, our fine-tuning recipe enables OpenVLA to successfully execute dexterous, high-frequency control tasks on a bimanual ALOHA robot and outperform other VLAs ($\pi_0$ and RDT-1B) fine-tuned using their default recipes, as well as strong imitation learning policies trained from scratch (Diffusion Policy and ACT) by up to 15% (absolute) in average success rate. We release code for OFT and pretrained model checkpoints at https://openvla-oft.github.io/.