VLCP (Vision Language Control Policy) is a training‑free robot manipulation approach that keeps a vision‑language model (VLM) frozen and uses it to generate short Python control functions. Unlike traditional methods that retry a fixed policy, VLCP rewrites the control code every K steps based on multi‑view RGB, proprioceptive state, and state delta, allowing failures to be corrected within the same episode. In a 57‑task MuJoCo/RoboVerse benchmark, VLCP achieves 35.1% pooled success versus 3.5% for a single‑query baseline, with a 27.3% within‑episode recovery rate on failed grasps and efficient token usage.
By Dhia Naouali, Minghan Wu, Claudia Wong, Abhinav Puthran, Omar G. Younis
RePO‑VLA is a recovery‑driven policy optimization framework for Vision‑Language‑Action models that separates success, recovery, and failure trajectories. It introduces Recovery‑Aware Initialization, a Progress‑Aware Semantic Value Function, and Value‑Conditioned Refinement to learn from both successful and failed rollouts, enabling the policy to prefer high‑progress actions without online failure detectors. The authors also present the Failure‑Recovery Benchmark (FRBench) and demonstrate that RePO‑VLA raises adversarial success rates from 20% to 75% on average in simulated and real‑world bimanual tasks.
By Weijia Liufu, Xiaoyu Guo, Ruiyi Chen, Jingzhi Liu, Kaidong Zhang, Xiwen Liang, Jianqi Lin, Dawei Sun, Yuze Wang, Rongtao Xu, Bingqian Lin, Bowen Yang, Tongtong Cao, Bowen Peng, Dongyu Zhang, Guangrun Wang, Min Wang, Liang Lin, Xiaodan Liang
The paper investigates a specific failure mode in vision‑language‑action models, termed instruction‑action binding, where models respond to language and vision separately but fail to combine them to select the correct action under counterfactual changes. Through behavioral analyses of fine‑tuned policies, the authors show that failed rollouts often preserve source behavior or switch to other demonstrated tasks, indicating that language is not ignored but mis‑bound. They propose Equivariant Counterfactual Training (ECT), which supplies counterfactual demonstrations and a paired loss to enforce correct action selection, achieving significant performance gains across simulated and real‑world benchmarks.
By Hung-Jen Chen, Yu-Hsun Hou, Yan-Hong Chen, Yan-Fu Chen, Binghua Cai, Min Sun, Chun-Yi Lee
arXiv:2606. 09630v1 Announce Type: cross Abstract: Vision-language-action (VLA) policies provide strong priors for language-conditioned manipulation, but remain brittle in off-nominal states requiring targeted recovery.
By Haodi Hu, Chung-Ta Huang, Jing Liu, Ye Wang, Kei Suzuki, Matthew Brand, Toshiaki Koike-Akino
The paper introduces FailBank, a four‑stage self‑evolving framework that transforms runtime feedback from safety shields into lasting policy improvements for vision‑language‑action (VLA) models. By using a counterfactual correction teacher, outcome‑aware admission, and guarded LoRA updates, FailBank converts useful shield proposals into corrective targets while preserving successful actions as anchors. Experiments on the VLA‑Arena benchmark show that FailBank boosts task success rates by up to 8.5 percentage points and reduces cumulative policy cost by up to 35.6%, outperforming both base policies and traditional runtime shielding.
By Mingyue Cui, Zheyuan Liu, Yihan Zhu, Zheyuan Zhang, Meng Jiang
arXiv:2607. 29169v1 Announce Type: cross Abstract: Vision-language-action (VLA) policies achieve strong performance in robotic manipulation but remain vulnerable to runtime disturbances that break the temporal alignment among visual observations, robot states, and executed actions.
By Wenda Yu, Tianshi Wang, Fengling Li, Xin Li, Jingjing Li, Lei Zhu
The paper investigates whether a single descriptive sentence can correct failures in frozen vision‑language‑action (VLA) policies. By labeling failures in the LIBERO‑90 benchmark, the authors identify four failure modes, finding that only execution misalignment—where the policy reaches the correct target but misjudges distance or force—can be remedied with a concise human‑generated sentence. This sentence, parsed by an LLM into a bias on the policy’s actions, successfully restores performance on both simulated and real‑world tasks, while automated language models fail to achieve similar gains.
By Owen Kwon, Pablo Ortega-Kral, Arthur Bucker, Jean Oh
arXiv:2606. 27755v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models enable instruction-driven robotic manipulation, but they inherit oversized language backbones from pretrained VLMs whose capacity far exceeds what is needed for short robotic instructions.
By Guoheng Sun, Kaixi Feng, Shwai He, Xiaochuan Gong, Yexiao He, Ziyao Wang, Zheyu Shen, Wanghao Ye, Ramana Rao Kompella, Gaowen Liu, Ang Li
ValueFormer introduces stage‑aware, continuous value labels for online failure detection in vision‑language‑action policies. The method uses a 3.5 M‑parameter causal transformer over a frozen DINOv3 backbone to emit per‑frame smooth Monte‑Carlo values for advantage estimation and sharp binary signals for mistake detection. In experiments, reusing teleoperation intervention flags boosts detection AP from 0.38 to 0.82 and flags 95 % of mistakes early; on a bimanual sandwich station, per‑frame post‑training weighting raises task completion from 70 % to 85 % and eliminates a repeat‑pick failure mode.
By Inkyu Sa, Konstantin Stulov, Rajat Bhageria
The paper investigates how small action errors evolve when using action chunking in behavioural cloning. By injecting errors at each state and observing their growth under open‑loop (no replanning) and closed‑loop (replanning) regimes, the authors classify states as contracting, expanding, or unresolved. Across twelve manipulation tasks, they find that stable states are rare, error amplification is common, and that short‑horizon fitting can overestimate long‑horizon propagation. Predictors trained on camera and proprioceptive data can recover open‑loop stability but only partially capture closed‑loop dynamics, indicating that standard imitation learning does not reliably produce policies that contract errors when perturbed.
By Aryan Goyal
The paper introduces Martingale GRPO (MART‑GRPO), a method for selectively truncating gradient contributions in vision‑language‑action (VLA) reinforcement learning. By defining stopping conditions based on success‑probability and policy‑score martingales, the approach removes gradient terms that would otherwise penalize potentially useful recovery actions while preserving the expectation of the GRPO estimator. Experiments on RoboTwin 2.0 with OpenVLA‑OFT demonstrate that MART‑GRPO yields higher success rates across multiple tasks compared to full‑trajectory GRPO.
By Zongyao Zhao, Cheng-Jie Zhao, Ruihang Jiang, Bing Zheng, Zhaolin Wang, Yuanwei Liu
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