How does AnchorQ work?
Real-world results
We evaluate three manipulation tasks on a Franka robot with a DROID setup, giving each RL method 100 online training episodes per task. Across 20 held-out configurations per task, AnchorQ outperforms frozen π0.5 and DSRL on all three tasks and is the only method to improve towel hanging over the base policy. Aggregate error bars show 95% Wilson confidence intervals.
Average error bars show 95% Wilson confidence intervals. p-values compare AnchorQ with each baseline using two-sided t-tests.
View data
Showing AnchorQ on Stack Cups.
Simulation Results
We compare π0.5, DSRL, and AnchorQ with adaptive particle tilting on two LIBERO-90 task suites. Each plot shows individual task success across 100 episodes followed by the average across tasks, with mean ±1 standard error over four training seeds.
- π0.5 zero-shot
- DSRL
- AnchorQ
Per-task success
LIBERO-90 task ID · hover or select a task for details
Average
All 15 tasks
Task and average error bars show ±1 SEM across four seeds. The average includes every task in the selected suite.
Watch the policies learn
DSRL
AnchorQ
For AnchorQ rollouts, the colored overlays identify the nearest anchor and its reference direction from the executed latent.