Zero-Shot Chain-of-Thought — Step-by-Step Reasoning

I will design and test a zero-shot chain-of-thought prompt for your task — using reasoning elicitation phrases calibrated to your specific LLM and task type, structuring the prompt to produce explicit intermediate reasoning steps before the final answer, and testing across your representative inputs to confirm CoT improves accuracy over direct prompting. Zero-shot CoT elicitation requires more than appending "think step by step" — different LLMs respond differently to different elicitation phrases, the positioning of the elicitation instruction within the prompt affects reasoning quality, and task-specific reasoning guidance (e.g. "identify the relevant quantities before calculating") produces significantly better reasoning than generic elicitation.

The design covers LLM-specific elicitation phrase selection, prompt positioning for reasoning elicitation, task-specific reasoning guidance, output parsing for final answer extraction, and accuracy comparison against your current direct-answer baseline.