Question 1

After fine-tuning a large language model (LLM) for generating legal documents, what is the most effective way to assess whether the fine-tuning has improved the model’s performance for this specific task?

Options :

A : Comparing the fine-tuned model’s output with that of a non-fine-tuned model on random text generation tasks.

B : Testing the fine-tuned model on a set of common, non-legal text generation tasks to measure general improvement.

C : Evaluating the model’s output against a benchmark dataset of legal documents that it has never seen before.

D : Measuring the speed at which the fine-tuned model generates text, regardless of content accuracy.

Answer: C

Question 2

You are testing the effectiveness of a multimodal AI model designed to predict stock market trends using financial news articles, social media sentiment, and historical stock prices. The model performs well during periods of market stability but shows significant accuracy drops during market volatility. What approach should you take to improve the model's performance during volatile periods?

Options :

A : Reduce the complexity of the model to avoid overfitting to stable market conditions.

B : Incorporate additional data sources such as real-time trading volumes and macroeconomic indicators to provide more context during volatile periods.

C : Focus training exclusively on periods of market stability to ensure consistent accuracy.

D : Apply data augmentation techniques to artificially increase the size of the dataset during stable periods.

Answer: B

Question 3

You are designing a generative AI system that needs to interpret and generate both textual descriptions and corresponding images. The system must integrate these diverse data types into a coherent model framework. Which of the following is the most effective approach for achieving this integration?

Options :

A : Convert all data types into a common format, such as text, before feeding them into the model.

B : Implement a multimodal transformer that processes text and image data together.

C : Use a pre-trained text model and extend it with additional layers to process images.

D : Use a separate pipeline for each data type and merge the outputs at the final stage.

Answer: B

Question 4

You are developing a generative AI system that needs to create text-based narratives based on a sequence of images. Which approach would best handle this multimodal task while ensuring accurate context understanding and efficient processing?

Options :

A : Use a GAN to generate synthetic images and a rule-based system to generate text.

B : Use a Vision Transformer (ViT) to encode the images and a GPT model fine-tuned on narrative generation to generate the text.

C : Use a CNN to encode the images and an RNN to generate the text.

D : Use a ResNet model to encode the images and a BERT model to generate the text.

Answer: B

Question 5

You are deploying a multimodal AI system that combines text, images, and audio to assist in emergency response decision-making. The system will be used by various agencies across different countries. Which approach will most effectively ensure that the AI system provides reliable and unbiased recommendations in diverse scenarios?

Options :

A : Using a single, high-quality dataset from a leading country with advanced emergency response systems.

B : Training the model exclusively on data from regions where it will be deployed.

C : Incorporating real-time feedback from users in different regions to continuously adapt the model.

D : Implementing strict usage guidelines to limit the system’s decision-making to only certain types of emergencies.

Answer: C