Policy Model
In robotics, a model that maps perceptual input directly to action output. It is a core component of reinforcement learning or imitation learning and must run in real time on the robot body.
Timeline (last 90 days)
The actual value of a small policy model that can run in real time on a robot body may exceed that of a large model requiring round-trip cloud inference, which would fail due to latency in dynamic environments.
In robotics, a policy model usually refers to a model that maps perceptual input directly to action output, and is a core component of reinforcement learning or imitation learning.
The denoising effect of the reward model propagates to the behavior of the final policy model, and a less noisy reward signal makes policy optimization more consistent with true human preferences.
All Facts (3)
The actual value of a small policy model that can run in real time on a robot body may exceed that of a large model requiring round-trip cloud inference, which would fail due to latency in dynamic environments.
50%UnverifiedIn robotics, a policy model usually refers to a model that maps perceptual input directly to action output, and is a core component of reinforcement learning or imitation learning.
50%UnverifiedThe denoising effect of the reward model propagates to the behavior of the final policy model, and a less noisy reward signal makes policy optimization more consistent with true human preferences.
50%