GenAI and LLM interview questions

Citations let users and evals check claims against retrieved IDs.

Lower it for extraction, classification, SQL, and customer-support facts.

Temperature shapes the full distribution

Fine-tuned facts are frozen until the next training run.

Parameter-efficient fine-tuning trains small adapter matrices instead of all weights.

Further training on a narrow dataset can erase general skills the base model had.

Fine-tune for house style, tool schemas, and refusal policy

Loaders ingest PDFs, HTML, and databases into Document objects.

Buffer memory stores raw turns, summary memory compresses old history, and vector memory retrieves past facts by similarity.

LangChain Expression Language composes steps as runnable units with standard invoke, batch, and stream methods.

The model returns a function name and JSON arguments instead of a user-facing answer.

Schemas tell the model required fields, types, and enums so arguments are usable.

The model may emit several function calls in one turn, for example weather in two cities.

Live APIs and databases supply ground truth the weights never had.

Offline evals run a frozen set of cases in CI.

Faithfulness asks whether claims are supported by retrieved context.

They reward n-gram overlap with a reference, but many correct phrasings exist.

Keep a labeled suite, run it on every prompt diff, and fail the build on score drops.

Jailbreaks target the model's safety policy, usually in the user turn.

Use a clear system policy, output classifiers, and refusal evals.