This paper presents tight bounds on the dimension of random projection for tensor product of vectors, achieving exponential improvement on sketch dimension compared to the prior work. Their method also enjoyes efficient computation. The reviewers found the work solid and of high significance.