COMPUTER-VISIONCURRENT2026-05-11

VISOR: A Vision-Language Model-based Test Oracle for Testing Robot

Prasun Saurabh, Pablo Valle, Aitor Arrieta, Shaukat Ali, Paolo Arcaini

Instead of hand-coding task-specific pass/fail checks or relying on humans to evaluate Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. behavior, VISOR uses vision-language models to automatically assess whether a Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. completed its Core ConceptsTaskThe job the robot is supposed to complete, such as pick-and-place, navigation, or drawer opening. correctly and with good quality just by watching video. This cuts Simulation & Sim-to-RealEvaluationMeasuring how well a robot system performs. time dramatically for testing Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. systems.

THE PROBLEM

This paper focuses on computer vision. VISOR applies VLMs (GPT, Gemini) as automated test oracles for Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Core ConceptsTaskThe job the robot is supposed to complete, such as pick-and-place, navigation, or drawer opening. Simulation & Sim-to-RealEvaluationMeasuring how well a robot system performs.. Given video of Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. behavior, it judges Core ConceptsTaskThe job the robot is supposed to complete, such as pick-and-place, navigation, or drawer opening. correctness and quality without symbolic hardcoding. The paper also attempts to quantify Modern Robot LearningVision-Language Model (VLM)A model that understands both images and text. uncertainty but finds it doesn't correlate well with actual correctness, limiting its use as a confidence Evaluation & ResearchMetricA numerical measure of performance.. 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 computer vision. Start here because it defines what success means and which assumptions the rest of the method inherits.

2

Core method

Instead of hand-coding task-specific pass/fail checks or relying on humans to evaluate Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. behavior, VISOR uses vision-language models to automatically assess whether a Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. completed its Core ConceptsTaskThe job the robot is supposed to complete, such as pick-and-place, navigation, or drawer opening. correctly and with good quality just by watching video. This cuts Simulation & Sim-to-RealEvaluationMeasuring how well a robot system performs. time dramatically for testing Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. systems. 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

Instead of hand-coding task-specific pass/fail checks or relying on humans to evaluate Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. behavior, VISOR uses vision-language models to automatically assess whether a Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. completed its Core ConceptsTaskThe job the robot is supposed to complete, such as pick-and-place, navigation, or drawer opening. correctly and with good quality just by watching video. This cuts Simulation & Sim-to-RealEvaluationMeasuring how well a robot system performs. time dramatically for testing Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. systems.

WHY DEVELOPERS SHOULD CARE

Instead of hand-coding task-specific pass/fail checks or relying on humans to evaluate Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. behavior, VISOR uses vision-language models to automatically assess whether a Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. completed its Core ConceptsTaskThe job the robot is supposed to complete, such as pick-and-place, navigation, or drawer opening. correctly and with good quality just by watching video. This cuts Simulation & Sim-to-RealEvaluationMeasuring how well a robot system performs. time dramatically for testing Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. systems.

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 computer vision 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