Sequential Hypothesis Testing for LLM Identification

This framework treats identifying a specific LLM from a known set as a two-stage optimization problem. It uses an outer probe-design stage to create model fingerprints and an inner sequential testing stage to minimize the evaluation budget needed to identify the active model.