Wait, that's not an option: LLMs Robustness with Incorrect Multiple-Choice Options

Gracjan Góral, Emilia Wiśnios, Piotr Sankowski +1

This work introduces a novel framework for evaluating LLMs' capacity to balance instruction-following with critical reasoning when presented with multiple-choice questions containi…