Module 3 Quiz: Autonomous AI Agents & Multi-Agent Systems - Deep Dive Track

Q1

What is the primary difference between a chatbot and an autonomous AI agent?

Conceptual

Chatbots use GPT-4 while agents use GPT-3.5

Agents can use external tools and take actions, chatbots only generate text

Chatbots are faster and cheaper to run

There is no meaningful difference

✓ Correct Answer: B

The key distinction is autonomy and tool use. Agents can reason about what actions to take, invoke external tools (search, APIs, calculators), and iteratively work toward goals. Chatbots primarily generate conversational text without external interactions.

Q2

In the ReAct pattern, what does "ReAct" stand for?

Definition

Re-activate and Continue

Reasoning and Acting

Recursive Action

Response and Action Cycle

✓ Correct Answer: B

ReAct = Reasoning + Acting. The pattern alternates between the agent thinking about what to do (reasoning) and actually doing it (acting with tools). This creates a loop: Thought → Action → Observation → Thought → ...

Q3

What is the correct order of steps in a ReAct agent's execution loop?

Process

Action → Thought → Observation → Final Answer

Thought → Action → Observation → (repeat or answer)

Observation → Action → Thought → Final Answer

Final Answer → Thought → Action

✓ Correct Answer: B

The ReAct loop follows: Thought (agent reasons) → Action (agent uses tool) → Observation (agent sees tool output) → repeat until the agent has enough information to provide a Final Answer.

Q4

What gives an AI agent "autonomy"?

Conceptual

The ability to decide which tools to use and when

Running on a dedicated server

Using the latest GPT model

Having a large context window

✓ Correct Answer: A

Autonomy comes from the agent's ability to make decisions about which actions to take, which tools to invoke, and when to stop. The LLM acts as the "reasoning engine" that determines the next step based on observations.

Q5

Which framework is best suited for building cyclic, stateful workflows with conditional branching?

Framework Selection

LangChain

LangGraph

CrewAI

AutoGPT

✓ Correct Answer: B

LangGraph is specifically designed for building complex, cyclic workflows with nodes, edges, and conditional logic. It excels at scenarios where you need loops, quality checks, and state management across multiple steps.

Q6

What is CrewAI's primary design philosophy?

Framework Concepts

Graph-based state machines

Role-based agent collaboration

Low-level control of LLM calls

Single-agent optimization

✓ Correct Answer: B

CrewAI focuses on role-based collaboration. You define agents with specific roles, goals, and backstories, then assign them tasks. The framework handles orchestration using sequential or hierarchical processes.

Q7

What distinguishes AutoGPT from LangChain-based agents?

Framework Comparison

AutoGPT is fully autonomous with long-running goals and memory

AutoGPT only works with GPT-4

LangChain is faster and cheaper

There is no difference

✓ Correct Answer: A

AutoGPT is designed for fully autonomous operation with long-term goals, persistent memory, and the ability to spawn sub-agents. LangChain agents are typically more focused and require tighter orchestration.

Q8

When selecting tools for an agent, what is the most important consideration?

Best Practice

Tools should be as complex as possible

Tool descriptions must be clear and unambiguous for the LLM

More tools always equals better performance

Tools should never return errors

✓ Correct Answer: B

The LLM selects tools based on their descriptions. Clear, concise, and unambiguous descriptions are critical. Poor descriptions lead to incorrect tool selection and agent failures.

Q9

In a multi-agent system, what is the most common pattern for agent communication?

Architecture

Direct telepathy between agent LLMs

Shared state or message passing through an orchestrator

Email-based communication

Agents never communicate directly

✓ Correct Answer: B

Agents typically communicate through shared state (LangGraph) or message passing coordinated by an orchestrator (CrewAI). Each agent's output becomes input for the next agent or is added to shared context.

Q10

Which coordination strategy is best for a research project requiring multiple specialized roles?

System Design

Sequential: Researcher → Writer → Critic

Parallel: All agents work simultaneously

Hierarchical: Manager delegates to sub-agents

Random: Agents act independently

✓ Correct Answer: A (or C)

Sequential coordination works best when tasks have clear dependencies (research must complete before writing). Hierarchical is also valid if complexity requires a manager agent to coordinate. Both answers would be acceptable in practice.

Q11

What is the primary benefit of agent specialization in multi-agent systems?

Benefits

Reduced API costs

Each agent can focus on a specific task with relevant tools and prompts

Faster execution time

Agents work in isolation without coordination

✓ Correct Answer: B

Specialization allows each agent to excel at its specific role. A research agent has search tools and is prompted for thoroughness. A writer agent focuses on clarity and structure. This division of labor improves overall quality.

Q12

What is task decomposition in the context of multi-agent systems?

Definition

Breaking complex problems into smaller sub-tasks that can be assigned to specialized agents

Deleting unnecessary tasks from the workflow

Combining multiple agents into one

Running the same task multiple times

✓ Correct Answer: A

Task decomposition is the process of breaking down a large, complex goal (e.g., "write a research report") into manageable sub-tasks (research, outline, write, edit, fact-check) that can be tackled by specialized agents.

Q13

Which memory type persists across application restarts?

Technical - Code

# Option A
memory = ConversationBufferMemory()

# Option B
memory = ConversationSummaryMemory(llm=llm)

# Option C
memory = Chroma(
    embedding_function=embeddings,
    persist_directory="./memory"
)

# Option D
memory = {"history": []}

Option A (ConversationBufferMemory)

Option B (ConversationSummaryMemory)

Option C (Chroma vector store)

Option D (Dictionary)

✓ Correct Answer: C

Chroma with persist_directory writes embeddings to disk, allowing memories to survive application restarts. Buffer and summary memory are in-memory only and lost on restart. Dictionaries are also ephemeral.

Q14

In LangGraph, what does the state schema define?

LangGraph Concepts

The structure of data passed between nodes in the graph

The physical location of the graph

The LLM provider to use

The visual appearance of the graph

✓ Correct Answer: A

The state schema (typically a TypedDict) defines the structure of shared data that flows through the graph. Each node can read from and write to this state, enabling coordination and information sharing.

Q15

Why is conversation summarization important for long-running agents?

Memory Management

It makes the agent sound more professional

It prevents context window overflow and reduces token costs

It's required by OpenAI's terms of service

It has no real benefit

✓ Correct Answer: B

Summarization compresses conversation history into key points, preventing the context window from filling up and drastically reducing token costs for long interactions. Without it, agents would exceed context limits or become prohibitively expensive.

Q16

What is the primary advantage of vector-based long-term memory?

Memory Systems

It's faster than all other memory types

Semantic search allows retrieving relevant memories even if keywords don't match

It requires no API calls

It works without embeddings

✓ Correct Answer: B

Vector memory uses semantic search via embeddings. You can query "agent frameworks" and retrieve memories about "LangGraph and CrewAI" even if those exact words weren't used. This is far more powerful than keyword matching.

Q17

What does this FastAPI endpoint do?

Technical - Code

@app.post("/agent/stream")
async def query_agent_stream(request: AgentRequest):
    return StreamingResponse(
        stream_agent_response(request.query),
        media_type="text/event-stream"
    )

Returns the complete agent response at once

Streams agent output in real-time using Server-Sent Events

Downloads a video stream

This code will not compile

✓ Correct Answer: B

This endpoint uses Server-Sent Events (SSE) to stream the agent's output in real-time. The client receives updates as the agent thinks and acts, providing a responsive user experience.

Q18

What is the best practice for handling agent tool failures?

Production Best Practice

Crash the application immediately

Return the error message to the agent so it can try a different approach

Hide all errors from the agent

Retry infinitely until it works

✓ Correct Answer: B

The agent should see tool errors so it can reason about alternatives. If a search fails, the agent might try a different query or use a different tool. Use handle_parsing_errors=True in LangChain to enable this.

Q19

Which technique is most effective for reducing agent API costs?

Cost Optimization

# Option A
llm = ChatOpenAI(model="gpt-4", temperature=0)

# Option B
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)

# Option C
agent.max_iterations = 100

# Option D
memory = ConversationBufferMemory()

Option A: Always use GPT-4

Option B: Use GPT-3.5-turbo for simpler tasks

Option C: Increase max iterations

Option D: Use buffer memory without summarization

✓ Correct Answer: B

Model routing (using GPT-3.5-turbo for simple tasks and GPT-4 for complex ones) can reduce costs by 10x. Other techniques: limit max_iterations, cache responses, use summarization, and compress prompts.

Q20

What is the most important consideration when deploying agents to production?

Production Deployment

Always use the largest context window available

Implement timeouts, rate limiting, monitoring, and error handling

Deploy without testing to move fast

Remove all logging for performance

✓ Correct Answer: B

Production agents need robust error handling: timeouts prevent infinite loops, rate limiting prevents abuse, monitoring tracks performance and costs, and comprehensive logging aids debugging. Never deploy agents without these safeguards.

Autonomous AI Agents & Multi-Agent Systems

Quiz Results