Return to Article Details Enhancing Zero-Shot Reasoning in Language Models Via Hybrid Instruction Marginalization Download Download PDF