New Method Stabilizes LLM Performance on Multiple-Choice Questions

Researchers propose Inclusion-of-Thoughts (IoT), a strategy that filters distracting answer choices to improve large language model accuracy on multiple-choice questions. The method reduces instability caused by plausible distractors across reasoning and knowledge benchmarks while enhancing model transparency.