ShieldVLA: Feasibility-Aware Safety Alignment for Vision-Language-Action Models

Hugging Face Daily Papers Papers
safety-alignment vision-language-action robotics ai-safety fine-tuning hamilton-jacobi

Summary

ShieldVLA proposes a safety-aligned fine-tuning framework for Vision-Language-Action models using Hamilton-Jacobi reachability to estimate safe operating regions, reducing safety costs by 57% and improving task success in robotics benchmarks.

Vision-Language-Action (VLA) models demonstrate strong generalization in robotic manipulation and navigation, but existing fine-tuning methods provide limited safety guarantees. Current approaches primarily rely on Lagrangian optimization that enforces safety through soft penalties on expected cumulative cost, often resulting in residual constraint violations or overly conservative behavior. Moreover, learning safety in visual domains is challenging due to the absence of dense per-step safety annotations. We propose ShieldVLA, a safety-aligned fine-tuning framework for VLA models based on Hamilton-Jacobi (HJ) reachability. ShieldVLA learns a model-free approximation of the HJ reachability value function directly from visual observations to estimate the safe operating region. The learned safety critic gates policy optimization by separating reward maximization within feasible regions from recovery near unsafe states, avoiding persistent reward-cost trade-offs. To enable scalable supervision in visual environments, we introduce rubric-based VLM safety scores that convert semantic safety feedback into structured critic targets without requiring manual cost labels. Across five navigation and manipulation benchmarks spanning multiple VLA backbones, ShieldVLA reduces cumulative safety cost by 57% on average and improves task success rate by +0.13 over SafeVLA.
Original Article
View Cached Full Text

Cached at: 09/22/26, 03:28 PM

Paper page - ShieldVLA: Feasibility-Aware Safety Alignment for Vision-Language-Action Models

Source: https://huggingface.co/papers/2609.13231

Abstract

Vision-Language-Action(VLA)modelsdemonstratestronggeneralizationinroboticmanipulationandnavigation,butexistingfine-tuningmethodsprovidelimitedsafetyguarantees.CurrentapproachesprimarilyrelyonLagrangianoptimizationthatenforcessafetythroughsoftpenaltiesonexpectedcumulativecost,oftenresultinginresidualconstraintviolationsoroverlyconservativebehavior.Moreover,learningsafetyinvisualdomainsischallengingduetotheabsenceofdenseper-stepsafetyannotations.WeproposeShieldVLA,asafety-alignedfine-tuningframeworkforVLAmodelsbasedonHamilton-Jacobi(HJ)reachability.ShieldVLAlearnsamodel-freeapproximationoftheHJreachabilityvaluefunctiondirectlyfromvisualobservationstoestimatethesafeoperatingregion.Thelearnedsafetycriticgatespolicyoptimizationbyseparatingrewardmaximizationwithinfeasibleregionsfromrecoverynearunsafestates,avoidingpersistentreward-costtrade-offs.Toenablescalablesupervisioninvisualenvironments,weintroducerubric-basedVLMsafetyscoresthatconvertsemanticsafetyfeedbackintostructuredcritictargetswithoutrequiringmanualcostlabels.AcrossfivenavigationandmanipulationbenchmarksspanningmultipleVLAbackbones,ShieldVLAreducescumulativesafetycostby57%onaverageandimprovestasksuccessrateby+0.13overSafeVLA.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2609\.13231

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.13231 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.13231 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.13231 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Seeing Before Colliding: Anticipatory Safe RL with Frozen Vision-Language Models

arXiv cs.LG

This paper presents VLM-Safe-RL, a framework that integrates frozen vision-language models into constrained MDP Lagrangian updates to provide anticipatory cost signals for safe reinforcement learning in high-speed visual control tasks. The method outperforms standard constraint-aware baselines on Safety-Gymnasium FormulaOne L2 and generalizes to held-out environments.

LabVLA: Grounding Vision-Language-Action Models in Scientific Laboratories

Hugging Face Daily Papers

LabVLA is a vision-language-action model for scientific laboratory automation, trained with a two-stage approach combining action token pretraining and flow matching. It achieves state-of-the-art success rates on the LabUtopia benchmark by leveraging simulated data to bridge the gap between household demonstrations and lab-specific tasks.