Short-term memory (conversation context) in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Short-term Memory in Conversation Context

Which of the following best describes the role of short-term memory in conversation context for an AI agent?

A It temporarily holds recent conversation details to maintain context during interaction.

B It stores all past conversations permanently for future reference.

C It deletes all conversation data immediately after each user input.

D It only stores user preferences without any conversation details.

💡 Hint

Think about how an AI keeps track of what was just said to respond properly.

Short-term memory in conversation context temporarily holds recent information to help the AI understand and respond appropriately during an ongoing interaction.

❓ Model Choice

intermediate

Choosing a Model for Short-term Memory in Dialogue

Which model architecture is best suited for capturing short-term memory in conversation context?

A K-Nearest Neighbors (KNN) for classification

B Recurrent Neural Network (RNN) with gated units like LSTM or GRU

C Convolutional Neural Network (CNN) for image recognition

D Feedforward Neural Network without recurrence

💡 Hint

Consider models designed to handle sequences and remember recent inputs.

RNNs with gated units like LSTM or GRU are designed to remember recent inputs, making them suitable for short-term memory in conversations.

❓ Metrics

advanced

Evaluating Short-term Memory Effectiveness

Which metric would best help evaluate how well an AI model retains short-term conversation context?

A BLEU score measuring translation quality

B Perplexity measuring prediction uncertainty on next tokens

C Recall measuring how many relevant recent context tokens are used in response

D Mean Squared Error measuring regression accuracy

💡 Hint

Think about measuring how much recent relevant information is captured in the output.

Recall measures how many relevant recent context tokens are included in the AI's response, indicating how well short-term memory is used.

🔧 Debug

advanced

Debugging Short-term Memory Loss in Conversation

An AI agent forgets the last user question immediately after responding. Which of the following is the most likely cause?

A The AI is using a large context window.

B The training data is too large.

C The model uses attention mechanisms correctly.

D The short-term memory buffer is cleared after each response.

💡 Hint

Consider what happens if the memory holding recent conversation is reset too soon.

If the short-term memory buffer is cleared after each response, the AI loses recent context and forgets the last user question.

❓ Predict Output

expert

Output of Short-term Memory Simulation Code

What is the output of this Python code simulating short-term memory with a fixed-size queue?

Agentic AI

from collections import deque

memory = deque(maxlen=3)
inputs = ['Hi', 'How are you?', 'What is AI?', 'Tell me a joke']
outputs = []
for inp in inputs:
    memory.append(inp)
    outputs.append(list(memory))
print(outputs)

A [['Hi'], ['Hi', 'How are you?'], ['Hi', 'How are you?', 'What is AI?'], ['How are you?', 'What is AI?', 'Tell me a joke']]

B [['Hi'], ['How are you?'], ['What is AI?'], ['Tell me a joke']]

C [['Hi'], ['Hi', 'How are you?'], ['Hi', 'How are you?', 'What is AI?'], ['Hi', 'How are you?', 'What is AI?', 'Tell me a joke']]

D [['Hi'], ['Hi', 'How are you?'], ['How are you?', 'What is AI?'], ['What is AI?', 'Tell me a joke']]

💡 Hint

Remember that deque with maxlen drops oldest items when full.

The deque holds up to 3 items. After adding the fourth input, the oldest ('Hi') is removed, so the last list contains the last three inputs.