問題文
A tester keeps seeing reasoning mistakes when asking a model to work out a large prioritization problem in one request. Which mitigation technique from the syllabus addresses that pattern most directly?
選択肢
- Convert the tables into cleaner readable formats, because the tester can remove the vague notation before the following prioritization step.
- Lower the sampling into narrower probability bands, because the tester can trim the random variation before the following prioritization step.
- Send the request into rival hosted engines, because the tester can compare the returned answers before the following prioritization step.
- Divide the prompt into smaller ordered pieces, because the tester can check the interim answer before the following prioritization step.