ACRO: Actor-Critic Rollout Orchestration
We introduce ACRO, an Actor--Critic rollout orchestration layer that makes these decisions for a fixed VLA.
ProofPaper ↗
Key points
- Executing a vision-language-action (VLA) policy involves deciding when to continue an attempt and when to intervene to improve its chances of success.
- A Critic learned from policy rollouts estimates success prospects to decide when to intervene and select a re-entry target along the executed trajectory.
- Trajectory Retraction shortcuts the visited path to return the robot to that target, where the same VLA resumes execution from a fresh observation.
- We evaluate ACRO with GR00T N1.7 and $π{0.5}$ across SimplerEnv, RoboCasa, and real-robot manipulation.
Sources (1)
- [1]ACRO: Actor-Critic Rollout OrchestrationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 06:55 AM
We introduce ACRO, an Actor--Critic rollout orchestration layer that makes these decisions for a fixed VLA.
Executing a vision-language-action (VLA) policy involves deciding when to continue an attempt and when to intervene to improve its chances of success.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 14, 2026Back on Track: Deployment-time Policy Improvement via World Model Planning
- Oct 14, 2026MorphAct: Robot Policies Morphing with the Worlds They Act In
- Oct 14, 2026G^3VLA: Geometric Inductive Bias for Vision-Language-Action Models
- Oct 13, 2026Breaking the Average: Overcoming Convergent Defaults in Robotic Manipulation via Diagnostic Chain-of-Thought and Structured Guidance
- Oct 13, 2026WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction
- Oct 13, 2026GreenWAM: A World Foundation Model for Robotics