1 paper
Morris de Haan, Philipp Hager
Despite the popularity of the two-tower model for unbiased learning to rank (ULTR) tasks, recent work suggests that it suffers from a major limitation that could lead to its collap…