GenAI and LLM interview questions

BERT stands for Bidirectional Encoder Representations from Transformers.

Classic GPT is a decoder-only transformer.

BERT is an encoder-only transformer.

GPT is trained as a language model that writes the next token, so generation is its native task.

A prompt is the text you give the model to steer what it does next.

A system prompt is a high-priority instruction that sets role, tone, tools, and hard rules.

Zero-shot prompting asks the model to perform a task with instructions only, no labeled examples.

Few-shot prompting includes a handful of input-output examples in the prompt.

Chain-of-thought prompting asks the model to reason step by step before the final answer.

In-context learning is the model adapting to a task from examples or instructions inside the prompt, without changing weights.

A user message is the human or application turn.

RAG retrieves relevant documents and stuffs them into the prompt so the LLM answers with that evidence.

Training data is frozen, incomplete, and not your internal corpus.

Retrieval is the step that selects the most relevant chunks for a query.

Grounding means the answer is tied to provided sources rather than invented facts.

A vector database stores embeddings and finds nearest neighbors at scale.

An index is the data structure that makes similarity search fast, such as HNSW or IVF.

ANN search returns vectors that are very close to the query without guaranteeing the exact closest set.

Hallucination is fluent text that is false, unsupported, or fabricated.

They fill gaps with statistically plausible tokens when evidence is missing or conflicting.