Translation validation aims to check the equivalence between an input program P and its translated counterpart P′. Traditional approaches focus on directly proving properties related to P and P′. Recently, we explored a new direction that makes translation validation more scalable and expressive by leveraging the translation records produced by the underlying optimizing compiler during the transformation from P to P′. In this talk, I will illustrate how compiler translation records can be used to improve the validation of various optimization passes in LLVM.