Vision-Language-Action Model, Robustness, Multi-modal Learning, Robot Manipulation

2026-04-11

AI Summary

This work addresses the limited robustness of existing vision-language-action (VLA) models under multimodal perturbations and the inherent trade-off between robustness and task accuracy in joint training. To resolve this, we propose STRONG-VLA, a two-stage decoupled fine-tuning framework: the first stage enhances robustness through curriculum learning with progressively harder multimodal perturbations, while the second stage recovers task execution capability on clean data, thereby avoiding conflicting optimization objectives. We introduce a comprehensive evaluation benchmark encompassing 28 perturbation types and demonstrate significant performance gains across multiple VLA models on LIBERO. Notably, OpenVLA achieves absolute success rate improvements of 12.60% and 7.77% under seen and unseen perturbations, respectively, with real-world robotic experiments further validating the efficacy of our approach.

Technology Category

• Computer Vision: Language and Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Despite their strong performance in embodied tasks, recent Vision-Language-Action (VLA) models remain highly fragile under multimodal perturbations, where visual corruption and linguistic noise jointly induce distribution shifts that degrade task-level execution. Existing robustness approaches typically rely on joint training with perturbed data, treating robustness as a static objective, which leads to conflicting optimization between robustness and task fidelity. In this work, we propose STRONG-VLA, a decoupled fine-tuning framework that explicitly separates robustness acquisition from task-aligned refinement. In Stage I, the model is exposed to a curriculum of multimodal perturbations with increasing difficulty, enabling progressive robustness learning under controlled distribution shifts. In Stage II, the model is re-aligned with clean task distributions to recover execution fidelity while preserving robustness. We further establish a comprehensive benchmark with 28 perturbation types spanning both textual and visual modalities, grounded in realistic sources of sensor noise, occlusion, and instruction corruption. Extensive experiments on the LIBERO benchmark show that STRONG-VLA consistently improves task success rates across multiple VLA architectures. On OpenVLA, our method achieves gains of up to 12.60% under seen perturbations and 7.77% under unseen perturbations. Notably, similar or larger improvements are observed on OpenVLA-OFT (+14.48% / +13.81%) and pi0 (+16.49% / +5.58%), demonstrating strong cross-architecture generalization. Real-world experiments on an AIRBOT robotic platform further validate its practical effectiveness. These results highlight the importance of decoupled optimization for multimodal robustness and establish STRONG-VLA as a simple yet principled framework for robust embodied control.

Problem

• Vision-Language-Action Model
• Robustness
• Multi-modal Learning
• Robot Manipulation
• Distribution Shift

Innovation

• Decoupled Fine-tuning
• Multimodal Robustness
• Vision-Language-Action Model
• Curriculum Perturbation Learning
• Embodied Control