AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

On Robustness of Vision-Language-Action Model against Multi-Modal Perturbations

Jianing Guo, Zhenhong Wu, Chang Tu, Yiyao Ma, Xiangqi Kong, Zhiqian Liu, Jiaming Ji, Shuning Zhang, Yuanpei Chen, Kai Chen, Qi Dou, Yaodong Yang and 4 more

ICLR 2026 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add RobustVLA input and output robust trainingpi0 · LIBERO, 17 perturbations averagesuccess rate (%): 62.6 -> 76.6Improved
add RobustVLA training on clean taskspi0 · LIBERO, no perturbationsuccess rate (%): 96.0 -> 95.5No clear effect
add BYOVLA visual robustness (VLM inpainting)pi0 · LIBERO, non-visual perturbationssuccess rate (%): +0.0% in non-visual modalities; +4.0% avg visualNo clear effect
add input regularization to output-only robust training (ablation)RobustVLA · LIBERO, 17 perturbations averagesuccess rate (%): 64.8 -> 76.6Improved
add output regularization to input-only robust training (ablation)RobustVLA · LIBERO, 17 perturbations averagesuccess rate (%): 71.7 -> 76.6Improved
add UCB-based perturbation selection (ablation)RobustVLA · LIBERO, 17 perturbations averagesuccess rate (%): 69.3 -> 76.6Improved
use random-perturbation domain randomization instead of UCB adversarial selection (ablation)pi0 fine-tuning · LIBERO, 17 perturbations averagesuccess rate (%): 62.6 -> 61.8No clear effect
output regularization alone on observation perturbations (ablation)RobustVLA w/o in · LIBERO, Dead Pixelsuccess rate (%): 20.8 -> 40.8Improved
use larger PGD action epsilon (0.06 vs 0.03) (ablation)RobustVLA · LIBERO, action noises averagesuccess rate (%): 53.3 -> 53.2No clear effect
apply RobustVLA with limited real-world demonstrationspi0 · Real-world FR5, 25 demossuccess rate (%): 37.5 -> 92.5Improved
apply RobustVLA with abundant demonstrationspi0 · Real-world FR5, 100 demossuccess rate (%): 65.0 -> 95.0Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy