LLM Basics u0026 Interview Prep: Essential Concepts Quiz Quiz

1. Identifying Generative AI

   Which of the following best describes a key difference between generative AI and traditional discriminative AI?

   1. A. Generative AI creates new content, while discriminative AI classifies existing data.
   2. B. Generative AI sorts data into categories, while discriminative AI invents data.
   3. C. Generative AI uses only labeled data, while discriminative AI uses unlabeled data.
   4. D. Generative AI can only translate text, discriminative AI can only summarize.
   5. E. Generative AI deletes irrelevant information, discriminative AI stores it.

2. Tokens in Language Models

   In the context of large language models, what is a 'token'?

   1. A. A mathematical equation performed by the model
   2. B. A unit of text, such as a word or part of a word, used as input or output
   3. C. A binary code required for model authentication
   4. D. An image used for training the model
   5. E. A predefined answer stored in the database

3. Understanding Prompt Engineering

   What is one main goal of prompt engineering in large language models?

   1. A. Formatting input in a way that guides the model toward producing the desired output
   2. B. Coding new neural networks from scratch
   3. C. Training the model with audio samples
   4. D. Compressing the model's size
   5. E. Deleting incorrect responses automatically

4. Retrieval Augmented Generation (RAG)

   How does Retrieval Augmented Generation (RAG) improve large language model outputs?

   1. A. By randomly generating answers based on training data
   2. B. By retrieving relevant information from external data sources to support responses
   3. C. By compressing the text outputs
   4. D. By removing all ambiguous words from answers
   5. E. By prioritizing speed over accuracy

5. Purpose of Chunking

   Why do we use chunking strategies in preparing data for large language models?

   1. A. To make data easier for hardware to access by splitting it into meaningful segments
   2. B. To translate data into multiple languages automatically
   3. C. To increase the randomness in outputs
   4. D. To remove all small words from the dataset
   5. E. To convert text to images for better understanding

6. Vector Embeddings Explained

   What is a vector embedding in the context of large language models?

   1. A. A fixed-length numeric representation of text capturing its meaning
   2. B. A picture used to summarize a sentence
   3. C. A database error code
   4. D. A direct copy of the original document
   5. E. An automatically generated graph of outputs

7. Vector Database Basics

   What is one key difference between a vector database and a traditional database?

   1. A. Vector databases store data as high-dimensional vectors for similarity search, while traditional databases use structured tables
   2. B. Vector databases only store images, traditional databases only store text
   3. C. Vector databases are less secure than traditional databases
   4. D. Traditional databases operate offline, vector databases require the internet
   5. E. Vector databases cannot be queried by users

8. Temperature Parameter Meaning

   In large language models, what does the temperature parameter control?

   1. A. The randomness or creativity of the generated output
   2. B. The server's operating temperature
   3. C. The number of tokens per answer
   4. D. The model's training speed
   5. E. The amount of memory allocated

9. Stopping Criteria in LLMs

   Which of the following is an example of a stopping criteria for a large language model?

   1. A. Setting a specific stop sequence, such as ' ', to indicate when the model should halt generation
   2. B. Instructing the model to use uppercase only
   3. C. Ordering results alphabetically
   4. D. Requesting images instead of text outputs
   5. E. Forcing the model to repeat the same sentence

10. Purpose of Fine-Tuning

    Why might someone fine-tune a large language model on their own data?

    1. A. To adapt the model so it performs better on specific tasks or domains relevant to the user
    2. B. To increase the number of hidden layers in the model
    3. C. To make the model train faster on new datasets only
    4. D. To transform text into audio output
    5. E. To disable all randomization in answers