問題文
A workflow has one step that requires long multi-step reasoning over conflicting documents and another step that classifies short strings into five fixed categories at high volume. How should model families be matched to these steps?
選択肢
- Assign a reasoning-capable model to the analysis step and a smaller, cheaper model to the high-volume classification step
- Use the smallest available model for both steps and add more few-shot examples to the analysis step, on the basis that example count substitutes for reasoning depth
- Choose the model family based on which one has the largest context window, for both steps
- Use the largest available model for both steps so that quality is uniform