Agent perception-reasoning-action loop in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding the role of perception in the agent loop

In the agent perception-reasoning-action loop, what is the primary role of the perception component?

A To gather information from the environment and convert it into a form the agent can understand

B To decide the best action based on the current knowledge

C To store past actions for future reference

D To execute the chosen action in the environment

💡 Hint

Think about how the agent first learns about its surroundings.

Perception is about sensing and interpreting data from the environment so the agent can understand what is happening around it.

❓ Predict Output

intermediate

Output of a simple agent loop step

Given the following Python code simulating one step of an agent loop, what is the printed output?

Agentic AI

class SimpleAgent:
    def __init__(self):
        self.state = 'idle'
    def perceive(self, data):
        return data.upper()
    def reason(self, info):
        if 'HELLO' in info:
            return 'greet'
        return 'wait'
    def act(self, action):
        if action == 'greet':
            return 'Hello, human!'
        return '...' 

agent = SimpleAgent()
data = 'hello world'
info = agent.perceive(data)
action = agent.reason(info)
result = agent.act(action)
print(result)

A ...

B greet

C hello world

D Hello, human!

💡 Hint

Follow the flow: perceive converts to uppercase, reason checks for 'HELLO', then act responds.

The agent perceives 'hello world' as 'HELLO WORLD', finds 'HELLO' in it, reasons to 'greet', and acts by returning 'Hello, human!'.

❓ Model Choice

advanced

Choosing the best reasoning model for an agent

Which model type is best suited for the reasoning component of an agent that must make decisions based on uncertain and changing environments?

A Probabilistic graphical model (e.g., Bayesian network)

B Deterministic rule-based system

C Static lookup table

D Simple linear regression

💡 Hint

Consider models that handle uncertainty and update beliefs.

Probabilistic graphical models can represent uncertainty and update decisions as new data arrives, making them ideal for dynamic environments.

❓ Hyperparameter

advanced

Hyperparameter affecting agent action speed

In an agent using reinforcement learning for action selection, which hyperparameter primarily controls how quickly the agent explores new actions versus exploiting known good actions?

A Learning rate (alpha)

B Discount factor (gamma)

C Exploration rate (epsilon)

D Batch size

💡 Hint

Think about the balance between trying new things and sticking to what works.

The exploration rate (epsilon) controls the probability of choosing a random action, encouraging exploration over exploitation.

🔧 Debug

expert

Debugging an agent loop causing infinite action repetition

Consider this agent loop snippet that causes the agent to repeat the same action endlessly. What is the most likely cause?

Agentic AI

class LoopAgent:
    def __init__(self):
        self.memory = []
    def perceive(self, env):
        return env.get_state()
    def reason(self, state):
        if state not in self.memory:
            self.memory.append(state)
            return 'explore'
        else:
            return 'explore'
    def act(self, action):
        return f'Action: {action}'

agent = LoopAgent()
env = type('Env', (), {'get_state': lambda self: 'state1'})()
for _ in range(3):
    s = agent.perceive(env)
    a = agent.reason(s)
    print(agent.act(a))

A The perceive method does not update the environment state

B The reason method returns 'explore' regardless of memory, causing repeated actions

C The memory list is cleared after each action

D The act method modifies the action incorrectly

💡 Hint

Look at the condition and returned action in the reason method.

The reason method returns 'explore' whether the state is new or known, so the agent never changes its action, causing repetition.