VLA-FAIL introduces a lightweight framework for detecting failures in vision-language-action models used in robotic manipulation. It combines two novel detectors: the first-layer Mahalanobis distance (LLMD) identifies out-of-distribution states by comparing token-wise deviations in last-layer features to training data, while the second, action chunk consistency (ACC), detects failures by spotting inconsistencies between consecutive action chunks in receding-horizon control. The authors also propose a threshold‑independent metric, “AUCPDT,” to evaluate detection accuracy and latency, and show through real‑world and simulation experiments that LLMD and ACC together provide reliable and early failure detection, outperforming more expensive baseline methods.
By Florian Seligmann, Emiliyan Gospodinov, Enes Ulas Dincer, Gerhard Neumann
The paper introduces Hide-and-Seek, a framework for detecting failures in Vision‑Language‑Action (VLA) models during robot execution. It treats failure detection as a coarsely supervised learning problem, using inter‑trajectory and intra‑trajectory contrastive objectives to localize failure‑indicative actions without step‑level annotations. Experiments on LIBERO, VLABench, and a real‑world robotic platform show that Hide‑and‑Seek achieves state‑of‑the‑art multi‑task failure detection performance across several VLA policies.
By Seongheon Park, Wendi Li, Changdae Oh, Samuel Yeh, Zsolt Kira, Michael Hagenow, Sharon Li
The paper introduces FailureSpot, a label‑efficient method for detecting failures at the timestamp level in vision‑language‑action (VLA) policies. It first generates weak supervision from unlabeled VLA action chunks by identifying abnormal patterns, then employs active learning to annotate only the most uncertain trajectories. Experiments on multiple VLA policies demonstrate improved performance for both timestamp‑level and trajectory‑level failure detection.
By Jie Ma, Zongxi Liu, Yi Zhu
FailBench is a new benchmark for robot failure detection, containing 2,197 manipulation attempts from 14 public sources, with 75% of failures occurring naturally. The study evaluates 13 vision‑language model (VLM) detectors, finding the best model achieves only 0.77 mean balanced accuracy, and that fine‑tuned failure detectors often underperform general‑purpose VLMs. Performance varies with visual evidence, excelling when object motion is observable but dropping to near chance on contact‑intensive assembly tasks, and input‑level cropping of outcome‑relevant regions improves the top detector by 2.4 percentage points.
By Zaruhi Navasardyan, Tatul Danielyan, Hrant Davtyan
arXiv:2608. 04510v1 Announce Type: cross Abstract: Diffusion-based vision-language-action (VLA) policies can generate plausible actions even when their predictions are weakly grounded in the visual and language evidence defining the task.
By Suhas Hegde, Jitendra Yasaswi Bharadwaj Katta
VLA-Scope is a two‑stage framework designed to predict failures in vision‑language‑action models under distribution shifts. The first stage detects out‑of‑distribution inputs and classifies their shift categories using pooled image and language representations. For OOD inputs, the second stage updates failure risk during execution by combining shift category, action‑prefix features, and execution progress, achieving a ROC‑AUC of 0.8497 after 60 actions and outperforming baseline methods.
By Kaiwen Zhu, Dongfang Liu, Liangkai Liu
Vision‑Language‑Action models (VLAs) often fail to follow language instructions because they rely on vision shortcuts learned from biased datasets. The authors introduce LIBERO‑CF, a counterfactual benchmark that tests language following by providing alternative task instructions for the same scene layouts, revealing widespread counterfactual failures. They propose Counterfactual Action Guidance (CAG), a dual‑branch inference scheme that amplifies language conditioning, which consistently improves language grounding and task success across diverse VLAs without requiring extra data or architectural changes.
By Yu Fang, Yuchun Feng, Dong Jing, Jiaqi Liu, Yue Yang, Zhenyu Wei, Daniel Szafir, Mingyu Ding
The paper introduces Configured Failure Trapping, a new backdoor attack targeting Vision‑Language‑Action (VLA) models that activates through subtle textual triggers and forces the robot to fail in a specific, controlled manner. It presents a data engine for generating high‑quality target trajectories, an automated evaluation suite, and two benchmarks—Trap‑LIBERO and Trap‑RoboTwin—covering four failure modes. The authors propose TrapVLA, a method that learns trigger‑induced action residuals to steer policies toward the desired failure behavior, demonstrating effectiveness in both simulation and real‑world robotic experiments while maintaining performance on clean data.
By Jun-Hui Liu, Kun-Yu Lin, Yi-Lin Wei, Xu-Han Chen, Yinghao Li, Zhuohao Li, Yuan-Ming Li, Qing Zhang, Xiaoyi Fan, Dongmei Jiang, Yan Li, Wei-Shi Zheng
The paper introduces BAS‑VLA, a task‑semantic action calibration framework for vision‑language‑action models that addresses two failure modes: unnecessary action drift under appearance changes and insufficient behavioral change under semantic alterations. BAS‑VLA uses a breaking‑centered calibration core and a selective evidence‑gated preserving auxiliary to maintain performance on clean and semantics‑preserving conditions while suppressing stale‑task behavior. Experiments on OpenPI‑pi0.5 and LIBERO‑Object Milk‑Swap show high success rates on clean and preserved tasks, a dramatic drop under target‑object swaps, and improved robustness to style shifts from 42% to 70% without harming clean performance.
By Shuaijun Liu, Feiyang You, Chengyu Wu, Shuyang Hao, Chenglong Zhang, Jingyao Cai, Xingwei Chen, Ningxin Su
arXiv:2604.16677v2 Announce Type: replace-cross
Abstract: Vision-language-action (VLA) models have emerged as generalist robotic controllers that map visual observations and language instructions to...
By Zongyao Lyu, Lingling Chen, William J. Beksi
arXiv:2606. 12299v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models provide a natural language interface to robot control, but the mapping from language to behavior is often brittle and unintuitive: semantically similar instructions can induce drastically different behaviors, while some capabilities may not be elicitable through prompting alone.
By Hyun Joe Jeong, Gokul Swamy, Andrea Bajcsy
arXiv:2606. 08508v1 Announce Type: cross Abstract: Generative robot policies fail unpredictably at deployment: they hesitate at critical moments, drift off-task, or commit to unrecoverable actions.
By Bingjia Huang, Xiangyu Li, Xiang Wang, Liang Mi, Zixu Hao, Weijun Wang, Hao Wu, Kun Li, Yunxin Liu, Ting Cao