Yuliang Zou: Label-Efficient Visual Understanding with Consistency Constraints. Virginia Tech, Blacksburg, VA, USA 2022