Scaling by Diversified Experience for Vision-Language-Action Models

2026-06-08

AI Summary

This work addresses the deployment challenges of vision-language-action (VLA) models in real-world settings, where tightly coupled high-level reasoning and low-level control lead to unstable policy optimization. To mitigate this, the authors propose an intent decoupling mechanism that disentangles control-relevant features from reasoning context, combined with a similarity-guided reinforcement learning training paradigm to stabilize policy updates and alleviate distributional shift. The approach integrates multimodal pretraining, diverse experience replay, and efficient fine-tuning, preserving core vision-language capabilities while substantially improving task success rates and out-of-distribution generalization. Experimental results demonstrate consistent superiority over existing methods on both real-world robotic tasks and multimodal benchmarks.

Abstract

Vision-Language-Action models face significant challenges in real-world deployment due to the entanglement of high-level reasoning with low-level control, and the instability of policy optimization. In this paper, we introduce SyVLA, a robust VLA model trained with diversified experiences. We propose an Intention Decoupling algorithm to isolate control-relevant features from reasoning contexts and a similar-sample guided RL pipeline to stabilize policy updates and mitigate distribution shift. Extensive experiments on real-world robotic tasks and multi-modal benchmarks demonstrate that SyVLA achieves superior task success rates and stronger out-of-distribution generalization compared to existing methods, while effectively preserving core vision-language capabilities. Codes and Datasets is released on \href{https://sy-vla.github.io/}{project page}.

Problem

• Vision-Language-Action models
• real-world deployment
• policy optimization instability
• reasoning-control entanglement
• distribution shift

Innovation

• Intention Decoupling
• Vision-Language-Action Models
• Similar-Sample Guided RL
• Distribution Shift Mitigation
• Out-of-Distribution Generalization