Self-Improving Vision-Language-Action Models with Data Generation via Residual RL

VLMRe-assign community