RESEARCHInvestigateNEXT 12 MONTHS
Every Wrong Answer Counts: Option-Level Psychometrics for LLM Multiple-Choice Benchmarks
arXiv cs.CL — Computation and Language
Factual evidence
What the source reports
Researchers introduce the LLM Nominal Response Model (LLM-NRM), a psychometric framework analyzing incorrect MCQ choices to evaluate model behavior.
Open source