Learning from Runtime Feedback through Failure-Bank Self-Evolution for Vision-Language-Action Models
A new framework, FailBank, is introduced to improve policy learning for vision-language-action models by using runtime feedback for persistent policy improvement.
Save an API key to vote.