MCQA LLMs Evaluations

Improving LLM First-Token Predictions in Multiple-Choice Question Answering via Prefilling Attack