Chat completions endpoint in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding the role of system messages in chat completions

In the chat completions endpoint, what is the primary purpose of the system message?

A To specify the output format of the AI's response

B To represent the user's input or question to the AI

C To store the AI's previous responses for reference

D To provide instructions or context that guide the AI's behavior throughout the conversation

💡 Hint

Think about which message sets the tone or rules for the AI's replies.

The system message is used to give the AI instructions or context that influence how it responds to user messages. It sets the behavior for the entire chat session.

❓ Predict Output

intermediate

Output of a chat completion request with multiple messages

Given the following chat messages sent to the chat completions endpoint, what will be the role and content of the AI's first response message?

Prompt Engineering / GenAI

messages = [
  {"role": "system", "content": "You are a helpful assistant."},
  {"role": "user", "content": "What is 2 + 2?"}
]

# The AI responds based on these messages.

A {"role": "assistant", "content": "2 + 2 equals 4."}

B {"role": "user", "content": "2 + 2 equals 4."}

C {"role": "system", "content": "2 + 2 equals 4."}

D {"role": "assistant", "content": "I don't understand the question."}

💡 Hint

The AI replies as the assistant role with an answer to the user's question.

The AI always responds with the role 'assistant'. Given the user asked 'What is 2 + 2?', the AI should answer correctly with '2 + 2 equals 4.'

❓ Model Choice

advanced

Choosing the best model for a chat completion task

You want to build a customer support chatbot that can handle complex questions and maintain context over multiple turns. Which model is best suited for this task?

A A large chat-optimized model trained for multi-turn conversations

B A small, fast model optimized for single-turn question answering

C A model specialized in image recognition

D A model designed only for text summarization

💡 Hint

Consider which model type is designed for conversations with context.

Large chat-optimized models are trained to understand and maintain context across multiple messages, making them ideal for chatbots handling complex dialogs.

❓ Hyperparameter

advanced

Effect of temperature parameter in chat completions

What is the effect of increasing the temperature parameter in a chat completions request?

A The AI's responses become shorter and more concise

B The AI's responses become more random and creative

C The AI ignores the system message instructions

D The AI always repeats the same response

💡 Hint

Think about how randomness affects AI text generation.

Higher temperature values increase randomness, making the AI's responses more varied and creative. Lower values make responses more focused and deterministic.

❓ Metrics

expert

Evaluating chat completion quality with perplexity

You have two chat completion models. Model A has a perplexity of 15 on your test set, and Model B has a perplexity of 30. What does this tell you about their performance?

A Both models perform equally well

B Model B predicts the test data better and is more confident in its responses

C Model A predicts the test data better and is more confident in its responses

D Perplexity measures response length, so Model B produces longer answers

💡 Hint

Lower perplexity means better prediction of the data.

Perplexity measures how well a model predicts a sample. Lower perplexity means the model is less 'surprised' by the data and generally performs better.