In the context of Generative AI (GenAI), various embedding models are used to represent textual data.
Which of the following best describes the difference between Word2Vec, BERT, and Sentence-BERT embedding models?

• A. Word2Vec creates static word embeddings, BERT generates dynamic embeddings based on context, and Sentence-BERT produces embeddings specifically optimized for sentence-level tasks like semantic similarity.
• B. Word2Vec captures both word and sentence meanings in a single vector space, BERT generates only word embeddings, and Sentence-BERT generates embeddings for entire documents.
• C. Word2Vec captures contextual relationships between words, while BERT and Sentence-BERT generate sentence-level embeddings based on the overall document length.
• D. Word2Vec uses a transformer architecture for embedding generation, whereas BERT and Sentence-BERT use neural networks to model context.

When optimizing the tuning process in IBM watsonx Tuning Studio, what is the main purpose of fine-tuning metering options?

• A. To monitor the effectiveness of early stopping techniques applied during training.
• B. To enforce stricter thresholds for model accuracy in order to ensure high precision.
• C. To limit the number of iterations allowed for model training.
• D. To track the computational cost and time for fine-tuning models based on resource consumption.

During prompt engineering for IBM Watsonx, you need to understand how the decoding process works when generating responses.
Which of the following best describes a high-level overview of the decoding process in generative AI?

• A. Decoding involves translating the input data into a format that the AI model can understand before generating an output.
• B. Decoding is the final step in training, where the AI model verifies the accuracy of its outputs against a predefined set of labels.
• C. Decoding is the process where the model generates a response token-by-token, choosing each token based on the probability distribution over all possible tokens.
• D. Decoding occurs only in reinforcement learning, where the model refines its responses based on user feedback over multiple generations.

You are using IBM watsonx's generative AI model to generate responses for a chatbot, and you want to ensure that the model stops generating text when it encounters a specific phrase like ":End of Response." Which of the following settings for stop sequences is most appropriate to achieve this goal?

• A. Set the stop sequence to "<|stop|>"
• B. Set the stop sequence to "STOP"
• C. Set the stop sequence to "End of Response"
• D. Set the stop sequence to "\ n\ n"

While developing a Retrieval-Augmented Generation (RAG) system using the transformers library, you want to improve the retrieval quality by ensuring that your queries and documents are represented in the same latent space for effective similarity matching.
Which of the following techniques would be the most appropriate to ensure this alignment between queries and documents?

• A. Fine-tune a transformer model on a document-query similarity task, so that both queries and documents are encoded into the same vector space for retrieval.
• B. Use a randomly initialized transformer model to encode both documents and queries for unbiased similarity calculation.
• C. Use a pre-trained BERT model to encode the documents and a pre-trained GPT model to encode the queries, ensuring diversity in embeddings.
• D. Use different transformer models for documents and queries, and normalize their embeddings to align them in the same latent space.