20A customer's lawyer filed a brief containing case citations our legal research assistant invented. What do you change in the product?▼hardNewHarveyMicrosoftGoogle◆ premiumThe failure every legal AI product fears. The interviewer wants a response that makes invented citations structurally impossible to present as real, not a promise to improve the model.Practise this question →Read answer →
08Why do language models hallucinate, and what can a product team actually do about it?▼mediumNewAnthropicGooglePerplexity○ sign inEveryone knows models make things up. The interviewer wants to hear why, which kinds of hallucination exist, and a layered set of product defences, without promising to eliminate it.Practise this question →Read answer →
12How would you measure hallucination in a feature that summarises customer calls?▼mediumNewSalesforceMicrosoftZomato◆ premium'Hallucination rate' sounds like one number. In practice it is three different failures, and the one that matters most is the hardest to see.Practise this question →Read answer →
36Model A answers everything, right 24% of the time. Model B declines half and is wrong far less. Which do we ship?▼hardNewOpenAIGoogle DeepMindPerplexity◆ premiumAccuracy alone rewards guessing. Set a penalty for wrong answers from what errors cost in your product, score both models under it, and the choice usually flips.Practise this question →Read answer →