Interactive Post-Training for Vision-Language-Action Models

TL;DR

RIPT-VLA is a reinforcement-learning-based post-training method that significantly improves vision-language-action models' success rates with minimal supervision and demonstrates strong generalization and robustness.

Contribution

It introduces a scalable, interactive post-training paradigm for VLA models that enhances performance with sparse rewards and minimal data, outperforming existing methods.

Findings

Achieved 97.5% success rate on OpenVLA-OFT model

Improved QueST model success by 21.2%

Enabled a 4% SFT model to reach 97% success within 15 iterations

Abstract

We introduce RIPT-VLA, a simple and scalable reinforcement-learning-based interactive post-training paradigm that fine-tunes pretrained Vision-Language-Action (VLA) models using only sparse binary success rewards. Existing VLA training pipelines rely heavily on offline expert demonstration data and supervised imitation, limiting their ability to adapt to new tasks and environments under low-data regimes. RIPT-VLA addresses this by enabling interactive post-training with a stable policy optimization algorithm based on dynamic rollout sampling and leave-one-out advantage estimation. RIPT-VLA has the following characteristics. First, it applies to various VLA models, resulting in an improvement on the lightweight QueST model by 21.2%, and the 7B OpenVLA-OFT model to an unprecedented 97.5% success rate. Second, it is computationally efficient and data-efficient: with only one demonstration, RIPT-VLA enables an unworkable SFT model (4%) to succeed with a 97% success rate within 15 iterations. Furthermore, we demonstrate that the policy learned by RIPT-VLA generalizes across different tasks and scenarios and is robust to the initial state context. These results highlight RIPT-VLA as a practical and effective paradigm for post-training VLA models through minimal supervision.