RECALL: Recovery Experience Collection for Active Lifelong Learning in Vision-Language-Action Models

arXiv·medium signal

Vision-Language-Action models are usually fine-tuned by passive imitation, collecting more demonstrations for tasks where the policy fails; RECALL instead actively collects recovery experiences for lifelong learning. The idea — learning from how to recover from failures rather than only from successful demonstrations — generalizes a useful principle to any agent that needs to improve on its own weak spots.