Robustness of VLAs
VLA: Vision-Language-Action Model
Overview
This field studies whether robots guided by vision-language-action models can follow instructions and act reliably when viewpoints, lighting, objects, sensor inputs, or robot execution differ from training. Training approaches include perturbation and counterfactual data, consistency objectives, and world-model supervision; inference-time methods use adaptation, monitoring, action correction, and failure recovery. Benchmarks such as LIBERO-Plus and RoboTwin test shifts in simulation, while real-robot evaluations expose hardware faults and contact-related failures; increasingly, evaluations measure safety and behavior over a rollout, not just task completion. A central challenge is improving performance under distribution shift without weakening clean-task performance, sensitivity to meaningful instructions, or safe behavior.
Latest papers
239 papers20 papers in the last four weeks, down 67% on the four weeks before. 0.2% of all new papers.
FAN: Foresight Action Normalization for Continual Adaptation of Vision-Language-Action Models
VLA-Scope: Shift-Aware Failure Prediction for Vision-Language-Action Models
Beyond Patch Removal: Persistent Adversarial Effects in Vision-Language-Action Policies
SAVLA: Symmetry-Aware Vision-Language-Action Models for Robotic Manipulation
Beyond Single-Axis Testing: Paired Evaluation of Compound Robustness in Vision-Language-Action Policies
Validating, Not Sampling: Region-Level Robustness of Vision-Language and Vision-Language-Action Models
Atomic Motion Coordinate for Language-Steerable and Force-Responsive Manipulation
Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models
HuRo: Robotizing Human Videos for Scalable VLA Pretraining
Time-Frequency Geometric Cross-Attention for Chunked Vision-Language-Action Models
VLA-Corrector: Stage-Aware Observable State Understanding for Prompt-Based Closed-Loop Recovery of Vision-Language-Action Policies
Geometry Conditioning in an Embodied SLM: Training Controls and Robustness Diagnostics in a 0.8B Hybrid Model
GloVLA: Let Geometry Move and Local VLA Interact for Robust Object-Centric Manipulation in Unstructured Environments
Where Success Breaks: Failure-Boundary Learning for Robust Vision-Language-Action Models
CR-VLA-Force: Learning Control-aware Compliance VLA Model for Robust Contact-rich Robotic Manipulation
LIBERO-RECOVER: Beyond Task Success Towards Failure Recovery in Robotic Manipulation Models
WISE: World-model-guided Imagination Scheduling for Efficient Post-training of Vision-Language-Action Models
Sensing Which Modality Matters: Evidence-Gated Regularization for Robust VLA Policies
What Can One Sentence Fix? A Failure Taxonomy and a Minimal Language Correction for Frozen VLA Policies
FLARE: A Failure-Aware Framework for Autonomous Correction and Recovery in Visual-Language Robotic Manipulation
perfect" data leaves them unable to recover from common execution errors, such as a missed grasp, a dropped object, or an unexpected collision. In this paper, we propose FLARE, a novel framework that endows VLAs with robust error recovery capabilities through a Retry" and Reset" paradigm. First, we introduce a Retry" mechanism by injecting perturbation and bridging segments that decouple robot pose from environment state into demonstrations, enabling the policy to autonomously handle execution deviations. Second, to address critical, state-breaking (OOD) failures, we introduce a Reset" pipeline. We leverage an MLLM for offline failure analysis to automatically identify OOD states from execution videos. This analysis enables the efficient, targeted collection of a small library of object-centric Reset" skills, which are trained to restore the environment to a task-valid state. Our full framework integrates these learned policies. At inference, an online MLLM monitor arbitrates between task execution and ``Reset" skills. Experiments on challenging, contact-rich manipulation tasks show our approach significantly improves task success and robustness.