Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.40002

How Much Should a VLA Trust Its Reasoning? Step-Level Guidance as a Measured Action-Value Function

Dabin Seo·Eunhyek Joa
View original source ↗Read source document ↗Export metadata ↓

Abstract

Reasoning vision-language-action (VLA) policies write a reasoning trace and then decode the action from it. A wrong trace yields a wrong action, and even a correct-looking trace can mislead the plan. Existing remedies ask whether the reasoning is right. We ask how strongly it should steer. CoC-CFG applies classifier-free guidance between the policy's context and the same context without the reasoning span. One weight per Euler step attenuates or amplifies the reasoning, and w=1 recovers the original policy. We treat the choice of this weight as a Markov decision process whose action values are measured, not bootstrapped. With the expert frozen and the noise fixed, applying a weight for one step and finishing the plan under the current controller returns that weight's value exactly. Q-adaptive guidance fits these values with small critics and improves the controller conservatively. In the Alpasim driving simulator it outperforms every fixed weight and online RL (SAC, GRPO) in open loop, and in closed loop it passes 262 of 300 scenes, against 252 for the original policy. On held-out frames of the Waymo end-to-end driving benchmark, its first three stages outperform every fixed weight, four adaptive-guidance methods, SAC and GRPO. On the hidden test set, a frozen Alpamayo 1.5 with rule-based camera and vehicle conversions and Q-adaptive guidance sets a new state of the art among published methods: 8.135159 rater feedback score, against 8.079 for the best published method.

Explore related subjects

cs.ROcs.CVcs.LG

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Dabin Seo, Eunhyek Joa. 2026-10-29. How Much Should a VLA Trust Its Reasoning? Step-Level Guidance as a Measured Action-Value Function. https://arxiv.org/abs/2610.40002

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Aligned Delta-Triplane Transformers as Occupancy World Models

Occupancy World Models (OWMs) aim to predict future 3D occupancy scenes from historical observations and future ego motions, providing a useful world simulation tool for autonomous driving. Existing methods usually rely on large networks to implicitly align multi-frame historical states and predict future occupancy in a full-state manner, which is costly and redundant because most scene regions remain unchanged over short time intervals. In this paper, we propose Aligned Delta-Triplane Transformer (ADTT), a compact 4D OWM that explicitly aligns historical triplanes with ego-motion compensation and predicts only future scene changes. Based on the aligned triplane prior, a query-conditioned Transformer predicts residual changes, where ego-motion queries guide controllable future motion and learnable external queries capture scene changes beyond ego motion. The predicted changes are added to the aligned prior and decoded into future occupancy scenes. Experiments on Occ3D-nus show that ADTT achieves state-of-the-art forecasting performance with fewer parameters and higher FPS. Our code is publicly available online: https://github.com/kyoran/ADTT.

cs.RO↗
arXiv2026

OmniVIC: A Self-Improving Variable Impedance Controller with Vision-Language In-Context Learning for Safe Robotic Manipulation

We present OmniVIC, a universal variable impedance controller (VIC) enhanced by a vision language model (VLM), which improves safety and adaptation in any contact-rich robotic manipulation task to enhance safe physical interaction. Traditional VIC have shown advantages when the robot physically interacts with the environment, but lack generalization in unseen, complex, and unstructured safe interactions in universal task scenarios involving contact or uncertainty. To this end, the proposed OmniVIC interprets task context derived reasoning from images and natural language and generates adaptive impedance parameters for a VIC controller. Specifically, the core of OmniVIC is a self-improving Retrieval-Augmented Generation(RAG) and in-context learning (ICL), where RAG retrieves relevant prior experiences from a structured memory bank to inform the controller about similar past tasks, and ICL leverages these retrieved examples and the prompt of current task to query the VLM for generating context-aware and adaptive impedance parameters for the current manipulation scenario. Therefore, a self-improved RAG and ICL guarantee OmniVIC works in universal task scenarios. The impedance parameter regulation is further informed by real-time force/torque feedback to ensure interaction forces remain within safe thresholds. We demonstrate that our method outperforms baselines on a suite of complex contact-rich tasks, both in simulation and on real-world robotic tasks, with improved success rates and reduced force violations. OmniVIC takes a step towards bridging high-level semantic reasoning and low-level compliant control, enabling safer and more generalizable manipulation. Overall, the average success rate increases from 27% (baseline) to 61.4% (OmniVIC).

cs.RO↗
arXiv2026

Safe Learning for Contact-Rich Robot Tasks: A Survey from Classical Learning-Based Methods to Safe Foundation Models

Contact-rich tasks pose significant challenges for robotic systems due to inherent uncertainty, complex dynamics, and the high risk of damage during interaction. Recent advances in learning-based control have shown great potential in enabling robots to acquire and generalize complex manipulation skills in such environments, but ensuring safety, both during exploration and execution, remains a critical bottleneck for reliable real-world deployment. This survey provides a comprehensive overview of safe learning-based methods for robot contact-rich tasks. We categorize existing approaches into two main domains: safe exploration and safe execution. We review key techniques, including constrained reinforcement learning, risk-sensitive optimization, uncertainty-aware modeling, control barrier functions, and model predictive safety shields, and highlight how these methods incorporate prior knowledge, task structure, and online adaptation to balance safety and efficiency. A particular emphasis of this survey is on how these safe learning principles extend to and interact with emerging robotic foundation models, especially vision-language models (VLMs) and vision-language-action models (VLAs), which unify perception, language, and control for contact-rich manipulation. We discuss both the new safety opportunities enabled by VLM/VLA-based methods, such as language-level specification of constraints and multimodal grounding of safety signals, and the amplified risks and evaluation challenges they introduce. Finally, we outline current limitations and promising future directions toward deploying reliable, safety-aligned, and foundation-model-enabled robots in complex contact-rich environments. More details and materials are available at our \href{ https://github.com/jack-sherman01/Awesome-Learning4Safe-Contact-rich-tasks}{Project GitHub Repository}.

cs.RO↗