Interview Practice Questions

What do you mean by statistics on user features?

How would you prevent the model from promoting heavily discounted products during a sale?

How are we going to convert it?

What kind of MLP would we use, and what would we train it on?

What embedding model would we use for categorical features?

For example, how would you one-hot encode the shape of a gemstone?

What is TF-IDF and how is it used in natural language processing?

Explain how the Word2Vec method works and what is the difference between CBOW and Skip-gram.

What is the main idea behind the BERT model and how does it differ from previous models?

Explain what self-attention is and how it differs from cross-attention.

When should GloVe be used instead of Word2Vec?

Compare temperature sampling and top-k sampling. In what scenarios would each be preferable?

What is beam search, and how does it differ from greedy approach?

Explain how BERT's pre-training works, including MLM and NSP.

When and why should fine-tuning be used for transformer models?

What is the role of RLHF (Reinforcement Learning from Human Feedback) in training language models?

Explain the difference between CBOW and Skip-gram in the context of word2vec.

When and why is it preferable to use GloVe over word2vec?

How does the self-attention mechanism work in transformers?

Explain how BERT works and how the next sentence prediction (NSP) is implemented.