Question 1

In neural networks, the vanishing gradient problem refers to what problem or issue?

Options :

A :

The problem of overfitting in neural networks, where the model performs well on the training data but poorly on new, unseen data.

B : The issue of gradients becoming too large during backpropagation, leading to unstable training.

C :

The problem of underfitting in neural networks, where the model fails to capture the underlying patterns in the data.

D :

The issue of gradients becoming too small during backpropagation, resulting in slow convergence or stagnation of the training process.

Answer: D

Question 2

A company is deploying a Large Language Model (LLM) to automate customer support across multiple channels (e.g., chat, email, and social media). During the deployment, they notice that the model's performance varies significantly across channels. What could be the most likely cause of this issue?

Options :

A : The LLM was not trained on multi-channel data

B : Insufficient hardware resources allocated to the model

C : The LLM is not configured to handle multiple languages

D : Differences in data distribution across channels

Answer: D

Question 3

You have trained two Large Language Models (LLMs) for sentiment analysis on customer reviews. Model A has a Cross-Entropy Loss of 0.25 and an F1-score of 0.82 on the validation set. Model B has a Cross-Entropy Loss of 0.30 but an F1-score of 0.87. Which model would you choose for deployment, and why?

Options :

A : Model B, because it has a higher loss, indicating it has learned more generalized features

B : Neither, because the difference in Cross-Entropy Loss is more significant than the difference in F1-score

C : Model B, because a higher F1-score indicates better balance between precision and recall

D : Model A, because lower Cross-Entropy Loss indicates better model confidence

Answer: C

Question 4

A legal team is using a Large Language Model (LLM) to draft contract clauses. However, they find that the generated clauses often lack the necessary legal specificity and include ambiguous language. What prompt engineering technique would best improve the clarity and precision of the generated contract clauses?

Options :

A : Using a generic prompt like "Draft a contract clause."

B : Including clear instructions in the prompt to use precise legal terminology and avoid ambiguity.

C : Lowering the temperature parameter to reduce creativity and variability in the output.

D : Increasing the length of the prompt by adding examples of poorly written clauses.

Answer: B

Question 5

You are tasked with fine-tuning a pre-trained Large Language Model (LLM) for a customer service chatbot. The chatbot must handle a wide variety of customer inquiries with high accuracy while being sensitive to variations in language use across different regions. Which approach would best meet these requirements?

Options :

A : Fine-tune the model with region-specific datasets and use data augmentation

B : Use a one-size-fits-all approach with the original model

C : Train a new model from scratch with the global dataset

D : Apply transfer learning without additional regional data

Answer: A