VLAFlow: A Unified Training Framework for Vision-Language-Action Models via Co-training and Future Latent Alignment figure
AlphaXiv 中文概览(可滚动查看)