CalVLA trains vision-language-action policies to signal likely task failure
Researchers introduce CalVLA, a method that shapes native confidence in vision-language-action policies using outcome supervision, letting a policy predict likely task failure from its own action probabilities.
Read the full story at arXiv cs.AI ↗
Timeline · 1 report
- 2026-11-03 05:00 · arXiv cs.AI
CalVLA: Shaping Native Confidence in Vision-Language-Action Policies with Outcome Supervision