LEARNINGCURRENT2026-06-16

Visual Verification Enables Inference-time Steering and Autonomous Policy Improvement

Mingtong Zhang, Dhruv Shah

This framework lets you take any pre-trained Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Core ConceptsPolicyThe rule or model that maps observations or states to actions. and improve it at runtime by having a visual verifier judge whether actions will succeed, steering the Core ConceptsPolicyThe rule or model that maps observations or states to actions. away from bad decisions without retraining. Better yet, the Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. can then self-improve offline by learning from its own verified successes—matching expert-level Robot LearningSample efficiencyHow quickly a method learns from each example or interaction. with zero human Safety & DeploymentInterventionA human or safety system stepping in during robot operation..

THE PROBLEM

This paper focuses on learning. VERITAS pairs a generalist Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Core ConceptsPolicyThe rule or model that maps observations or states to actions. (generator) with a visual verifier that evaluates Core ConceptsActionA command the robot sends to its motors, controller, or low-level system. feasibility at Evaluation & ResearchInference timeHow long the model takes to produce an output.. The verifier steers Core ConceptsPolicyThe rule or model that maps observations or states to actions. Core ConceptsExecutionActually carrying out planned or predicted actions on the robot. toward viable actions, then verified rollouts provide self-supervision for offline Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task.. Results show inference-time verification boosts performance without additional Robot LearningTrainingThe process of fitting a model using data or experience., and self-improved policies match expert Imitation & Reinforcement LearningDemonstrationAn example of a task being done correctly, often by a human. efficiency. Read the paper by tracking the Core ConceptsTaskThe job the robot is supposed to complete, such as pick-and-place, navigation, or drawer opening. definition, the Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. or data assumptions, and the evidence that supports the claimed improvement.

HOW IT WORKS

1

Task framing

The paper frames the work as learning. Start here because it defines what success means and which assumptions the rest of the method inherits.

2

Core method

This framework lets you take any pre-trained Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Core ConceptsPolicyThe rule or model that maps observations or states to actions. and improve it at runtime by having a visual verifier judge whether actions will succeed, steering the Core ConceptsPolicyThe rule or model that maps observations or states to actions. away from bad decisions without retraining. Better yet, the Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. can then self-improve offline by learning from its own verified successes—matching expert-level Robot LearningSample efficiencyHow quickly a method learns from each example or interaction. with zero human Safety & DeploymentInterventionA human or safety system stepping in during robot operation.. When reading the method section, identify the inputs, the learned or engineered representation, and the Core ConceptsActionA command the robot sends to its motors, controller, or low-level system. or prediction produced by the system.

3

Data and supervision

For robotics work, the data story is part of the method: check whether the system depends on Imitation & Reinforcement LearningTeleoperation (teleop)A human remotely controlling the robot, often to collect demonstrations., Simulation & Sim-to-RealSimulationA virtual environment where robots can be trained or tested., internet video, human labels, or Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. rollouts.

4

Evaluation evidence

The paper should be judged through its Simulation & Sim-to-RealEvaluationMeasuring how well a robot system performs. protocol: what data is used, what Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. or simulator is tested, and which Evaluation & ResearchBaselineA reference method used for comparison. comparisons support the claim. Look for the gap between the headline result and the Simulation & Sim-to-RealDeploymentPutting the trained system on a real robot. setting you would actually care about.

KEY RESULTS

Main contributionConceptual contribution

This framework lets you take any pre-trained Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Core ConceptsPolicyThe rule or model that maps observations or states to actions. and improve it at runtime by having a visual verifier judge whether actions will succeed, steering the Core ConceptsPolicyThe rule or model that maps observations or states to actions. away from bad decisions without retraining. Better yet, the Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. can then self-improve offline by learning from its own verified successes—matching expert-level Robot LearningSample efficiencyHow quickly a method learns from each example or interaction. with zero human Safety & DeploymentInterventionA human or safety system stepping in during robot operation..

WHY DEVELOPERS SHOULD CARE

This framework lets you take any pre-trained Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Core ConceptsPolicyThe rule or model that maps observations or states to actions. and improve it at runtime by having a visual verifier judge whether actions will succeed, steering the Core ConceptsPolicyThe rule or model that maps observations or states to actions. away from bad decisions without retraining. Better yet, the Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. can then self-improve offline by learning from its own verified successes—matching expert-level Robot LearningSample efficiencyHow quickly a method learns from each example or interaction. with zero human Safety & DeploymentInterventionA human or safety system stepping in during robot operation..

LIMITATIONS

The main limitation to check is whether the claimed behavior holds outside the paper's reported setup. That means testing across different Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. embodiments, scenes, objects, and data distributions.

WHAT COMES NEXT

The practical next step is independent reproduction with clear baselines, ablations, and stress tests. For a developer, the useful follow-up is to map the paper's learning assumptions onto a concrete Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. stack, then test the smallest version of the method that could run end to end.

RELATED PAPERS