A learned Bernoulli continuation policy that decides per action chunk whether to keep executing or re-plan improves success of frozen VLA policies on RoboTwin 2.0, LIBERO, and a real robot.
Proceedings of the AAAI Conference on Artificial Intelligence , volume=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Continue or Replan? Bernoulli-Continuation Policy Learning for Adaptive Horizon Execution
A learned Bernoulli continuation policy that decides per action chunk whether to keep executing or re-plan improves success of frozen VLA policies on RoboTwin 2.0, LIBERO, and a real robot.