Rewind-IL adds a zero-shot failure detector based on action-chunk discrepancy plus a checkpoint-based state respawn mechanism to make imitation-learned robot policies more reliable during long tasks.
Scaling Verification Can Be More Effective than Scaling Policy Learning for Vision-Language-Action Alignment,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Rewind-IL: Online Failure Detection and State Respawning for Imitation Learning
Rewind-IL adds a zero-shot failure detector based on action-chunk discrepancy plus a checkpoint-based state respawn mechanism to make imitation-learned robot policies more reliable during long tasks.