35Users keep asking to see how our assistant reached its answer. Should the product show the model's reasoning?▼hardNewAnthropicOpenAIMistral AI◆ premiumReasoning traces often leave out what drove the answer, and explanations make people accept wrong answers more. Show progress and evidence, and decide who gets the full trace.Practise this question →Read answer →
38How would you evaluate a deep research feature that writes ten-page reports with citations?▼hardNewPerplexityOpenAI◆ premiumShort-answer benchmarks say little about ten-page reports. Check each statement against the source it cites, score coverage against an expert outline, test balance on contested topics and measure whether users act without redoing the work.Practise this question →Read answer →