IMITATION-LEARNINGCURRENT2026-07-28

HiFi-UMI: Learning Deployable Manipulation Policies from High-Fidelity UMI Data Alone

Yuteng Wei, Jinming Ma, Jiawei Wang, Weitao Zhou, Yushen Zuo, Ke Rui, Minglei Li, Jinhao Zhang, Zhikang Pan, Xiang Wang, Haoran Jia, Huan Du, Zicheng Zeng, Jun Ma, Guiyu Qin, Di Zhang, Xiaofei Li

You can train Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies entirely from high-quality human demonstrations without any Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. by building better data-capture rigs (±3mm accuracy, synchronized multi-camera stereo-inertial Navigation & LocomotionSLAMSimultaneous Localization and Mapping.). The authors show policies trained only on HiFi-UMI data match or exceed Imitation & Reinforcement LearningTeleoperation (teleop)A human remotely controlling the robot, often to collect demonstrations. baselines across multiple Modern Robot LearningVision-Language-Action model (VLA)A model that takes images and language as input and outputs robot actions. architectures, and they're releasing 2,000 hours of this data for the community.

THE PROBLEM

This paper focuses on Imitation & Reinforcement LearningImitation Learning (IL)Teaching a robot by showing it examples of how to do a task.. You can train Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies entirely from high-quality human demonstrations without any Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. by building better data-capture rigs (±3mm accuracy, synchronized multi-camera stereo-inertial Navigation & LocomotionSLAMSimultaneous Localization and Mapping.). The authors show policies trained only on HiFi-UMI data match or exceed Imitation & Reinforcement LearningTeleoperation (teleop)A human remotely controlling the robot, often to collect demonstrations. baselines across multiple Modern Robot LearningVision-Language-Action model (VLA)A model that takes images and language as input and outputs robot actions. architectures, and they're releasing 2,000 hours of this data for the community. 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 Imitation & Reinforcement LearningImitation Learning (IL)Teaching a robot by showing it examples of how to do a task.. Start here because it defines what success means and which assumptions the rest of the method inherits.

2

Core method

You can train Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies entirely from high-quality human demonstrations without any Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. by building better data-capture rigs (±3mm accuracy, synchronized multi-camera stereo-inertial Navigation & LocomotionSLAMSimultaneous Localization and Mapping.). The authors show policies trained only on HiFi-UMI data match or exceed Imitation & Reinforcement LearningTeleoperation (teleop)A human remotely controlling the robot, often to collect demonstrations. baselines across multiple Modern Robot LearningVision-Language-Action model (VLA)A model that takes images and language as input and outputs robot actions. architectures, and they're releasing 2,000 hours of this data for the community. 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

You can train Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies entirely from high-quality human demonstrations without any Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. by building better data-capture rigs (±3mm accuracy, synchronized multi-camera stereo-inertial Navigation & LocomotionSLAMSimultaneous Localization and Mapping.). The authors show policies trained only on HiFi-UMI data match or exceed Imitation & Reinforcement LearningTeleoperation (teleop)A human remotely controlling the robot, often to collect demonstrations. baselines across multiple Modern Robot LearningVision-Language-Action model (VLA)A model that takes images and language as input and outputs robot actions. architectures, and they're releasing 2,000 hours of this data for the community.

WHY DEVELOPERS SHOULD CARE

You can train Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies entirely from high-quality human demonstrations without any Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. by building better data-capture rigs (±3mm accuracy, synchronized multi-camera stereo-inertial Navigation & LocomotionSLAMSimultaneous Localization and Mapping.). The authors show policies trained only on HiFi-UMI data match or exceed Imitation & Reinforcement LearningTeleoperation (teleop)A human remotely controlling the robot, often to collect demonstrations. baselines across multiple Modern Robot LearningVision-Language-Action model (VLA)A model that takes images and language as input and outputs robot actions. architectures, and they're releasing 2,000 hours of this data for the community.

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 Imitation & Reinforcement LearningImitation Learning (IL)Teaching a robot by showing it examples of how to do a task. 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