Numbered steps is a prompt engineering technique where the instructions provided to the language model are broken down into a detailed, step-by-step sequence with each step numbered. For example: 1. Identify all dates in the text 2. Extract just the year from each date 3. Replace each date with only the year 4. Remove any duplicate years 5. Sort the years in ascending chronological order Breaking down instructions into small, unambiguous steps makes it much easier for the language model to follow them correctly. Numbering the steps also helps the model understand that the instructions should be followed sequentially. Compared to verbose paragraphs, numbered steps tend to produce more accurate outputs from language models. The increased specificity and clarity of each sub-task results in better adherence to the overall prompt. See an example in [[example of steps for text analysis]]. [[few-shot prompting]] < [[Hands-on LLMs]]/[[5 Prompting]] > [[chain of thought prompting]]