InterviewStack.io LogoInterviewStack.io

Generative AI and Large Language Models Questions

The capabilities and behavior of modern generative and large language models. Covers how LLMs are pretrained, in-context learning and few-shot prompting, generative model families (autoregressive, diffusion), context windows, and tokenization and sampling. Emphasizes understanding what generative models can and cannot do and how they differ from discriminative ML.

EasyTechnical
79 practiced

What is the difference between prompting an LLM and fine-tuning it to change its behavior? In what scenarios is prompting a sufficient first approach, and when does it become insufficient, requiring fine-tuning instead?

HardTechnical
102 practiced

You must choose an approach for a production chatbot: (A) a supervised model fine-tuned on conversation logs, (B) retrieval plus a reranker, or (C) RL fine-tuning with human feedback (RLHF). Compare these on safety, response quality, data requirements, compute cost, and monitoring needs. Which would you choose for a first production release, and why?

EasyTechnical
87 practiced

Define 'hallucination' in the context of large generative models. What concrete forms can it take, and why do these occur from a training and objective-function perspective? At a high level, what categories of mitigation exist?

HardTechnical
94 practiced

You must decide between two third-party LLM options for a knowledge assistant: a faster, cheaper model with slightly lower factual accuracy, versus a slower, costlier model with better factuality. How would you evaluate and choose, and how might you combine both to meet product goals?

MediumTechnical
96 practiced

Explain the differences between zero-shot, one-shot, few-shot, and in-context learning in LLMs. Describe scenarios where each is preferred, and when you would reach for fine-tuning instead of relying on in-context capabilities.

That is every published Generative AI and Large Language Models question for Technical Product Manager so far. Browse the other topics in this category, or practice this one interactively.