REINFORCEMENT-LEARNINGCURRENT2026-06-10

Critic Architecture Matters: Dual vs. Unified Critics for Humanoid Loco-Manipulation

Mehmet Turan Yardımcı

When Robot LearningTrainingThe process of fitting a model using data or experience. humanoid robots to walk while manipulating objects, using separate critic networks for Navigation & LocomotionLocomotionMovement of the robot body through space, like walking, rolling, or running. and Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. tasks makes policies 3.5× faster and 2× more efficient than using a single critic. This is critical for developers Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. pre-trained Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies with Imitation & Reinforcement LearningReinforcement Learning (RL)Teaching a robot through trial and error using rewards.: a unified critic actually suppresses the learned behavior by creating competing gradients, so architectural choice matters more than Imitation & Reinforcement LearningRewardA score that tells the robot how well it is doing. engineering.

THE PROBLEM

This paper focuses on Imitation & Reinforcement LearningReinforcement Learning (RL)Teaching a robot through trial and error using rewards.. When Robot LearningTrainingThe process of fitting a model using data or experience. humanoid robots to walk while manipulating objects, using separate critic networks for Navigation & LocomotionLocomotionMovement of the robot body through space, like walking, rolling, or running. and Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. tasks makes policies 3.5× faster and 2× more efficient than using a single critic. This is critical for developers Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. pre-trained Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies with Imitation & Reinforcement LearningReinforcement Learning (RL)Teaching a robot through trial and error using rewards.: a unified critic actually suppresses the learned behavior by creating competing gradients, so architectural choice matters more than Imitation & Reinforcement LearningRewardA score that tells the robot how well it is doing. engineering. 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 LearningReinforcement Learning (RL)Teaching a robot through trial and error using rewards.. Start here because it defines what success means and which assumptions the rest of the method inherits.

2

Core method

When Robot LearningTrainingThe process of fitting a model using data or experience. humanoid robots to walk while manipulating objects, using separate critic networks for Navigation & LocomotionLocomotionMovement of the robot body through space, like walking, rolling, or running. and Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. tasks makes policies 3.5× faster and 2× more efficient than using a single critic. This is critical for developers Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. pre-trained Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies with Imitation & Reinforcement LearningReinforcement Learning (RL)Teaching a robot through trial and error using rewards.: a unified critic actually suppresses the learned behavior by creating competing gradients, so architectural choice matters more than Imitation & Reinforcement LearningRewardA score that tells the robot how well it is doing. engineering. 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

When Robot LearningTrainingThe process of fitting a model using data or experience. humanoid robots to walk while manipulating objects, using separate critic networks for Navigation & LocomotionLocomotionMovement of the robot body through space, like walking, rolling, or running. and Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. tasks makes policies 3.5× faster and 2× more efficient than using a single critic. This is critical for developers Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. pre-trained Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies with Imitation & Reinforcement LearningReinforcement Learning (RL)Teaching a robot through trial and error using rewards.: a unified critic actually suppresses the learned behavior by creating competing gradients, so architectural choice matters more than Imitation & Reinforcement LearningRewardA score that tells the robot how well it is doing. engineering.

WHY DEVELOPERS SHOULD CARE

When Robot LearningTrainingThe process of fitting a model using data or experience. humanoid robots to walk while manipulating objects, using separate critic networks for Navigation & LocomotionLocomotionMovement of the robot body through space, like walking, rolling, or running. and Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. tasks makes policies 3.5× faster and 2× more efficient than using a single critic. This is critical for developers Modern Robot LearningFine-tuningTaking a pretrained model and adapting it to a specific robot or task. pre-trained Manipulation & TasksManipulationUsing a robot arm or hand to move or interact with objects. policies with Imitation & Reinforcement LearningReinforcement Learning (RL)Teaching a robot through trial and error using rewards.: a unified critic actually suppresses the learned behavior by creating competing gradients, so architectural choice matters more than Imitation & Reinforcement LearningRewardA score that tells the robot how well it is doing. engineering.

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 LearningReinforcement Learning (RL)Teaching a robot through trial and error using rewards. 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