In item response theory (IRT) modeling for categorical data, two response processes are typically characterized: the cumulative process and the unfolding process. Most of IRT models, also called dominance models, deal with a cumulative response process, where the probability of a positive response is a monotonic function of the relevant parameters on the continuum. The unfolding IRT model finds an ideal-point of person location on the underlying continuum, creating a single-peaked and symmetric item response function curve. Recent research indicated the potential use of unfolding models for examining rater preference and accuracy in scoring tasks -- unfolding models can reveal more information than dominance models in reflecting individual differences among human raters. This talk will introduce the unfolding IRT models and illustrate its applications in examining rater-mediated assessments.