問題文
A data scientist writes a prompt that asks a language model to extract three fields from an invoice and return them. Which option correctly identifies the two practices that make the output easier to use downstream? (Select two.)
選択肢
- State the output format explicitly, so the response can be parsed by the next step without guessing.
- Include the whole invoice archive in every prompt, so the model has the maximum amount of context available.
- Ask the model to also decide whether the invoice should be paid, so that one single call ends up covering both the extraction and the payment decision at the same time.
- Tell the model what to do when a field is absent, so a missing field produces a defined value rather than an invented one.
- Keep the prompt as short as possible, because short prompts are more accurate.