AI AI Engineer: NLP and Large Language Models 2

Question 1

What is 'hallucination' in the context of LLMs?

Accepted Answer

The model producing plausible-sounding but factually incorrect or fabricated information

Answer

The model generating extremely long outputs

Answer

The model refusing to answer sensitive questions

Answer

The model outputting garbled text due to tokenization errors

Question 2

What is the context window in a large language model?

Accepted Answer

The maximum number of tokens the model can process in a single input/output sequence

Answer

The UI window showing model outputs

Answer

The time window used for model training

Answer

The layer of attention that focuses on the current word

Question 3

Which parameter-efficient fine-tuning technique adds low-rank decomposition matrices to model layers instead of updating all weights?

Accepted Answer

LoRA (Low-Rank Adaptation)

Answer

Full fine-tuning

Answer

RLHF (Reinforcement Learning from Human Feedback)

Answer

Knowledge distillation

Question 4

What is 'zero-shot prompting' when using an LLM?

Accepted Answer

Asking the model to perform a task without providing any examples in the prompt

Answer

Prompting the model with zero tokens

Answer

Training the model on zero labeled examples

Answer

Using a model that has never been fine-tuned

Question 5

What is 'embedding' in NLP, as used by transformer models?

Accepted Answer

A dense, fixed-size vector representation of a token or text that captures semantic meaning

Answer

Storing model weights in a database

Answer

The process of tokenizing input text

Answer

A technique for compressing the attention matrix

Question 6

What is RLHF (Reinforcement Learning from Human Feedback) used for in LLM development?

Accepted Answer

Aligning LLM outputs with human preferences and reducing harmful outputs

Answer

Speeding up pretraining

Answer

Expanding the model's vocabulary

Answer

Compressing the model for edge deployment