1 paper · 1 filter
Haihao Lin, Xiangsheng Huang, Xiao Yang +7
Action-supervised fine-tuning of vision-language-action (VLA) policies fits demonstrations effectively but constrains only the directions that change predicted actions, leaving vis…