Conversation management in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Context Windows in Conversation Management

In AI conversation systems, what is the main purpose of a context window?

A To store and use recent conversation history to maintain context

B To limit the number of users interacting with the AI simultaneously

C To encrypt user data for privacy during conversations

D To speed up the AI's response time by skipping processing

💡 Hint

Think about how the AI remembers what was said before.

The context window holds recent conversation turns so the AI can understand and respond appropriately, maintaining a natural flow.

❓ Predict Output

intermediate

Output of Conversation Turn Tracking Code

What is the output of this Python code that tracks conversation turns?

Prompt Engineering / GenAI

turns = ['Hi', 'Hello', 'How are you?', 'I am fine']
last_turn = turns[-1]
print(f'Last user input: {last_turn}')

A Last user input: How are you?

B Last user input: I am fine

C Last user input: Hello

D Last user input: Hi

💡 Hint

Look at how negative indexing works in Python lists.

Using turns[-1] accesses the last item in the list, which is 'I am fine'.

❓ Model Choice

advanced

Choosing a Model for Long Conversation Contexts

You want to build a chatbot that remembers details from a long conversation (over 1000 words). Which model type is best suited?

A A recurrent neural network (RNN) with gated units like LSTM or GRU

B A small feedforward neural network with no memory

C A simple linear regression model

D A k-nearest neighbors model

💡 Hint

Think about models that can remember sequences over time.

RNNs with LSTM or GRU units are designed to remember information over long sequences, making them suitable for long conversations.

❓ Metrics

advanced

Evaluating Conversation Quality with Metrics

Which metric best measures how well a conversation AI keeps track of user intent over multiple turns?

A Mean squared error measuring numeric prediction error

B BLEU score measuring word overlap

C Perplexity measuring language model uncertainty

D Slot filling accuracy measuring correct extraction of user details

💡 Hint

Focus on metrics related to understanding user information.

Slot filling accuracy measures how well the AI extracts and remembers user details, which is key for tracking intent.

🔧 Debug

expert

Debugging Conversation State Update Code

What error does this code raise when updating conversation state?

state = {'topic': 'weather', 'mood': 'happy'}
update = {'mood': 'curious', 'location': 'park'}
state.update(update['location'])
print(state)

A SyntaxError due to missing colon

B KeyError: 'location'

C ValueError: dictionary update sequence element #0 has length 1; 2 is required

D No error, prints updated state with new location

💡 Hint

Check what type is passed to the update() method.

The update() method expects a dictionary or iterable of key-value pairs, but update['location'] is a string, causing a ValueError: dictionary update sequence element #0 has length 1; 2 is required.