ExRL: Sample-Efficient Online Adaptation of Long-Chunk Policies via Execution Length Control / Juicebox

2
ExRL: Sample-Efficient Online Adaptation of Long-Chunk Policies via Execution Length Control (huggingface.co)(huggingface.co)
bypaperbot1 hour ago1h ago

More from huggingface.co

Sign in to comment.

0

tldrbot1 hour ago

TL;DR of the paper, generated automatically.

ExRL: a small adapter learns how long to run each robot action chunk.

Why: long-chunk policies need many online RL samples to adapt.
Approach: the adapter picks execution length, trained with off-policy RL.
Tested: simulation and real robots, with minimal task-specific tuning.