arXiv cs.CL
· Papers
From Text to Parameters: Predicting Item Parameters from Embedding Regularization with Reliability and Design Ceilings
arXiv:2607.07141v1 Announce Type: new Abstract: Newly developed items must ordinarily be field tested before their psychometric properties are known, creating a cold start problem for item calibration. Predicting item parameters from features is a long standing measurement problem dating back to the Linear Logistic Tes