Removing unnecessary steps reduces processing time and improves the efficiency of the workflow.
Incorrect! Try again.
21A logistics platform uses separate agents for route planning, inventory checking, and delivery scheduling. What makes this system a multi-agent system rather than a single modular program?
Introduction to multi-agent systems
Medium
A.All components use the same database and programming language
B.Each agent acts autonomously while interacting toward shared goals
C.All decisions are produced by one centralized prediction model
D.Each module runs sequentially within one fixed control flow
Correct Answer: Each agent acts autonomously while interacting toward shared goals
Explanation:
A multi-agent system contains autonomous entities that make decisions and interact or coordinate to achieve individual or shared objectives.
Incorrect! Try again.
22Three warehouse agents choose tasks independently. Two agents repeatedly select the same package while another package remains unassigned. Which capability is most directly missing?
Introduction to multi-agent systems
Medium
A.A coordination mechanism for resolving task conflicts
B.A larger context window for every warehouse agent
C.A faster database for storing completed delivery records
D.A shared language model for generating package labels
Correct Answer: A coordination mechanism for resolving task conflicts
Explanation:
The duplicate assignments indicate that agents lack a mechanism for coordinating choices and resolving conflicts over shared tasks.
Incorrect! Try again.
23In a planner-executor architecture, a report-generation system creates an outline, assigns research actions, and later revises the outline when a source is unavailable. Which component should perform the revision?
Planner-executor architectures
Medium
A.The monitoring dashboard, because it displays task status
B.The communication layer, because it formats agent messages
C.The executor, because it stores every generated document
D.The planner, because it updates the task sequence
Correct Answer: The planner, because it updates the task sequence
Explanation:
The planner owns decomposition and sequencing, so it should revise the plan when execution conditions invalidate an earlier step.
Incorrect! Try again.
24An executor reports that a required customer record does not exist. What is the most appropriate next action in a robust planner-executor system?
Planner-executor architectures
Medium
A.Allow the executor to redefine the business objective permanently
B.Return the result so the planner can replan or terminate
C.Repeat the same action indefinitely without changing the plan
D.Mark the entire workflow successful and suppress the error
Correct Answer: Return the result so the planner can replan or terminate
Explanation:
Execution feedback should return to the planner, which can choose an alternative step, request information, or end the workflow safely.
Incorrect! Try again.
25A diagnosis-support workflow includes agents that analyze images, summarize patient history, and check treatment guidelines. Which design best supports effective collaboration?
Collaborative agents
Medium
A.Require all agents to use identical prompts and identical tools
B.Let every agent produce a final diagnosis without exchanging results
C.Combine specialized outputs through a defined synthesis process
D.Have one agent perform all tasks while the others remain idle
Correct Answer: Combine specialized outputs through a defined synthesis process
Explanation:
Specialized agents create value when their complementary findings are integrated through an explicit synthesis or decision process.
Incorrect! Try again.
26Two research agents return conflicting estimates from equally credible sources. Which response best preserves collaborative reliability?
Collaborative agents
Medium
A.Select the estimate returned first and discard the other
B.Repeat both estimates until the agents produce matching values
C.Ask a reviewer agent to compare evidence and report uncertainty
D.Average the estimates without examining their assumptions
Correct Answer: Ask a reviewer agent to compare evidence and report uncertainty
Explanation:
A reviewer can assess evidence, identify differing assumptions, and communicate uncertainty instead of resolving disagreement arbitrarily.
Incorrect! Try again.
27Agents developed by different teams exchange task updates, but failures occur because each team uses different field names and status values. What is the most suitable remedy?
Inter-agent communication
Medium
A.Define a shared message schema and communication protocol
B.Store all agent prompts in a single text document
C.Replace asynchronous messages with longer natural-language messages
D.Increase the number of messages sent by each agent
Correct Answer: Define a shared message schema and communication protocol
Explanation:
A shared schema and protocol establish consistent fields, meanings, and status conventions across independently developed agents.
Incorrect! Try again.
28A monitoring agent publishes equipment alerts, and several maintenance agents must react independently. Which communication model is most appropriate?
Inter-agent communication
Medium
A.Direct polling of every agent through one shared prompt
B.Sequential file transfer between pairs of maintenance agents
C.Publish-subscribe messaging with alerts organized by topic
D.Permanent locking of the equipment database during each alert
Correct Answer: Publish-subscribe messaging with alerts organized by topic
Explanation:
Publish-subscribe communication lets multiple interested agents receive the same event without tightly coupling the sender to each receiver.
Incorrect! Try again.
29A coordinator must delegate a translation task requiring legal expertise and completion within ten minutes. Which policy is most appropriate?
Task delegation
Medium
A.Assign it to the agent with the longest operating history
B.Assign it randomly to reduce the coordinator's computation
C.Match required capability and deadline against agent availability
D.Send it to every agent and accept all returned translations
Correct Answer: Match required capability and deadline against agent availability
Explanation:
Effective delegation considers task requirements, agent capabilities, current workload, and timing constraints.
Incorrect! Try again.
30Agent A can finish a task in 6 minutes with quality score 0.8, while Agent B needs 10 minutes with quality score 0.95. The deadline is 8 minutes. If meeting the deadline is mandatory, which delegation is valid?
Task delegation
Medium
A.Delegate to Agent B and ignore the deadline constraint
B.Delegate to Agent A because it satisfies the deadline
C.Delegate to Agent B because it has the highest quality
D.Delegate to both and wait for both results
Correct Answer: Delegate to Agent A because it satisfies the deadline
Explanation:
Agent A is the valid choice because it completes the task within the mandatory 8-minute deadline, whereas Agent B does not.
Incorrect! Try again.
31Four regional agents optimize delivery routes locally, but their routes compete for the same limited loading dock. What additional step is needed to obtain a feasible global solution?
Distributed problem solving
Medium
A.Increase each region's route length by the same amount
B.Coordinate local plans using the shared dock constraint
C.Prevent regional agents from evaluating local delivery data
D.Choose the route produced by the largest regional agent
Correct Answer: Coordinate local plans using the shared dock constraint
Explanation:
Local solutions may conflict over shared resources, so the agents must reconcile their plans using the global dock-capacity constraint.
Incorrect! Try again.
32A problem can be divided into five independent simulations whose results are combined at the end. Which strategy is likely to reduce completion time most effectively?
Distributed problem solving
Medium
A.Run the simulations sequentially in order of agent identifier
B.Run the simulations in parallel and aggregate their results
C.Duplicate each simulation on every agent before aggregation
D.Delay all simulations until a central agent predicts their outputs
Correct Answer: Run the simulations in parallel and aggregate their results
Explanation:
Independent subtasks can be executed concurrently, after which their outputs can be combined into the final result.
Incorrect! Try again.
33A customer-support workflow always passes a request through classification, retrieval, response drafting, and compliance review in that order. Which orchestration pattern best fits this workflow?
Orchestration patterns
Medium
A.A peer-to-peer debate with unrestricted message exchange
B.A sequential pipeline with explicit stage transitions
C.A broadcast pattern with no dependency between stages
D.A competitive auction based only on execution cost
Correct Answer: A sequential pipeline with explicit stage transitions
Explanation:
A sequential pipeline is appropriate when each stage depends on the output of the preceding stage and the order is stable.
Incorrect! Try again.
34An analytics workflow sends the same dataset to three specialized agents and then combines their independent findings. Which orchestration pattern is being used?
Orchestration patterns
Medium
A.A single-agent loop with recursive planning
B.A token-passing ring without result merging
C.Fan-out followed by fan-in aggregation
D.A strict chain with one active branch
Correct Answer: Fan-out followed by fan-in aggregation
Explanation:
The work fans out to multiple agents in parallel, and their outputs fan in to an aggregation step.
Incorrect! Try again.
35Several agents need exclusive access to a shared machine. Which mechanism most directly prevents two agents from using it simultaneously?
Agent coordination mechanisms
Medium
A.A distributed lock or lease for the machine
B.A voting system for evaluating completed outputs
C.A larger prompt containing the machine's manual
D.A broadcast channel for announcing general capabilities
Correct Answer: A distributed lock or lease for the machine
Explanation:
A distributed lock or time-limited lease enforces mutual exclusion for a resource that only one agent may use at a time.
Incorrect! Try again.
36A team of agents must select one of three proposed plans, and no central coordinator is available. Which mechanism can produce a collective choice?
Agent coordination mechanisms
Medium
A.Consensus or voting over the proposed plans
B.Random deletion of two plans by the first agent
C.Exclusive database access for the proposing agents
D.Independent execution of every plan without comparison
Correct Answer: Consensus or voting over the proposed plans
Explanation:
Consensus or voting enables decentralized agents to evaluate alternatives and converge on a collective decision.
Incorrect! Try again.
37In an invoice workflow, an extraction agent reads invoice fields, a policy agent checks limits, and a payment agent issues funds. Which control is most important before payment?
Collaborative enterprise workflows
Medium
A.A validated approval event containing the policy decision
B.A copy of every message produced by the extraction agent
C.A shared visual theme across all agent interfaces
D.A second payment request generated by the payment agent
Correct Answer: A validated approval event containing the policy decision
Explanation:
The payment agent should act only after receiving a verifiable approval that confirms required policy checks have passed.
Incorrect! Try again.
38A software-delivery system has developer, reviewer, and release-manager agents. Which permission model best follows role-based design?
Role-based agents
Medium
A.Allow each role only the tools required for its responsibilities
B.Give every role full production access to avoid workflow delays
C.Base permissions only on the length of an agent's response
D.Rotate all permissions randomly after each completed task
Correct Answer: Allow each role only the tools required for its responsibilities
Explanation:
Role-based design should apply least privilege, granting each agent only the capabilities needed to perform its assigned responsibilities.
Incorrect! Try again.
39A multi-agent service slows down when one coordinator handles every message from 500 worker agents. Which change most directly addresses the bottleneck?
Scalable AI collaboration strategies
Medium
A.Increase the length of messages exchanged by all workers
B.Introduce hierarchical coordinators and partition the workers
C.Require all workers to send status messages more frequently
D.Move every task to the existing central coordinator
Correct Answer: Introduce hierarchical coordinators and partition the workers
Explanation:
Hierarchical coordination distributes communication and scheduling load, reducing pressure on a single central coordinator.
Incorrect! Try again.
40A workflow has three stages with average times of 2, 12, and 3 seconds. The stages must remain sequential, but the middle stage can use additional agents. Which optimization is most likely to improve throughput?
Workflow optimization for multi-agent systems
Medium
A.Delay the 3-second stage until several workflows accumulate
B.Parallelize or replicate processing at the 12-second stage
C.Add more agents only to the 2-second stage
D.Increase logging at every stage without changing execution
Correct Answer: Parallelize or replicate processing at the 12-second stage
Explanation:
The 12-second stage is the primary bottleneck, so increasing its processing capacity is most likely to improve overall throughput.
Incorrect! Try again.
41A multi-agent system has agents with locally optimal policies, but its global performance is worse than that of a centralized controller. Which explanation best accounts for this outcome?
Introduction to multi-agent systems
Hard
A.Local rewards are misaligned with the system-level objective
Correct Answer: Local rewards are misaligned with the system-level objective
Explanation:
Optimizing local objectives does not guarantee a globally optimal outcome. Misaligned rewards can create externalities, duplicated work, or competition that reduces total system performance.
Incorrect! Try again.
42Agents and repeatedly choose whether to share costly information. Sharing benefits both agents only when both participate, but withholding yields a short-term individual advantage. Which design most directly promotes sustained cooperation?
Introduction to multi-agent systems
Hard
A.Replace communication with independently sampled actions
B.Allow agents to optimize only their immediate rewards
C.Hide each agent's actions until all interactions finish
D.Use repeated interactions with reputation and future penalties
Correct Answer: Use repeated interactions with reputation and future penalties
Explanation:
Reputation and future consequences change the long-term payoff structure. They make opportunistic withholding costly and can stabilize cooperation across repeated interactions.
Incorrect! Try again.
43A planner creates a valid ten-step plan, but after step four the environment changes and invalidates a key precondition for step seven. What should a robust planner-executor architecture do?
Planner-executor architectures
Hard
A.Restart the original plan without incorporating the new observation
B.Detect the violated precondition and replan from the observed state
C.Continue execution because the plan was valid when generated
D.Ask the executor to reinterpret the failed precondition as satisfied
Correct Answer: Detect the violated precondition and replan from the observed state
Explanation:
Execution monitoring should compare observed conditions with plan assumptions. Once a required precondition becomes false, replanning from the current state avoids propagating an invalid plan.
Incorrect! Try again.
44A planner delegates a database migration to an executor. The executor times out, but the planner cannot determine whether the migration committed. Which interface property is most important for safe recovery?
Planner-executor architectures
Hard
A.Randomized retries with different migration parameters
B.Idempotent execution with queryable operation status
C.Higher token limits for the planner's internal context
D.A longer natural-language description of the migration
Correct Answer: Idempotent execution with queryable operation status
Explanation:
Queryable status resolves uncertainty about completion, while idempotency makes retries safe. Together they prevent duplicate or conflicting side effects after ambiguous failures.
Incorrect! Try again.
45Three diagnostic agents use the same model, training data, and retrieval source. A majority vote among them shows little improvement over one agent. Which change is most likely to improve collective accuracy?
Collaborative agents
Hard
A.Increase the number of identical agents using the same prompt
B.Require every agent to copy the first agent's conclusion
C.Remove disagreement before agents inspect the available evidence
D.Introduce agents with diverse evidence sources and reasoning roles
Correct Answer: Introduce agents with diverse evidence sources and reasoning roles
Explanation:
Majority voting helps mainly when errors are not strongly correlated. Diverse sources and roles reduce correlated failure modes and make agreement more informative.
Incorrect! Try again.
46Two agents propose incompatible solutions supported by different evidence. Which resolution policy best preserves useful disagreement while still producing a decision?
Collaborative agents
Hard
A.Force both agents to adopt the proposal with greater verbosity
B.Average the textual responses without examining their evidence
C.Use an adjudicator that evaluates claims against shared criteria
D.Select whichever proposal was submitted to the system first
Correct Answer: Use an adjudicator that evaluates claims against shared criteria
Explanation:
An adjudicator can compare evidence, assumptions, and constraints under explicit criteria. This resolves conflict without discarding minority evidence merely because it is less popular or arrived later.
Incorrect! Try again.
47In an asynchronous agent network, messages may be delayed, duplicated, or delivered out of order. Which message design best prevents an older update from overwriting a newer one?
Inter-agent communication
Hard
A.Increase message length and repeat every instruction twice
B.Attach entity versions and idempotency keys to each update
C.Use agent names as the only identifiers in all messages
D.Assume arrival order always matches the sender's creation order
Correct Answer: Attach entity versions and idempotency keys to each update
Explanation:
Versions allow receivers to reject stale writes, while idempotency keys identify duplicates. These controls address ordering and repeated-delivery failures independently.
Incorrect! Try again.
48An agent sends the message Approve account 1842 to another agent, but the receiver interprets it as permission to activate the account rather than permission to continue review. Which protocol improvement most directly removes this ambiguity?
Inter-agent communication
Hard
A.Increase the receiver's autonomy over downstream account actions
B.Compress the message to reduce network transmission latency
C.Send the same natural-language message through two channels
D.Use typed performatives with explicit action schemas and states
Correct Answer: Use typed performatives with explicit action schemas and states
Explanation:
Typed performatives distinguish intentions such as authorize, request, propose, or acknowledge. Explicit schemas also define the target action and permitted state transition.
Incorrect! Try again.
49A coordinator must assign task to one of three agents. Their success probabilities and execution costs are , , and . If utility is , which agent should receive the task?
Task delegation
Hard
A.The third agent, because its utility is
B.The second agent, because its utility is
C.The first agent, because its utility is
D.No agent, because every execution cost is positive
Correct Answer: The third agent, because its utility is
Explanation:
The utilities are , , and . Under the stated objective, the third agent has the highest expected utility.
Incorrect! Try again.
50A manager agent delegates a compliance check to a specialist that reports high confidence but cannot provide evidence. The action is irreversible and high impact. What is the best delegation policy?
Task delegation
Hard
A.Repeat the same request until the confidence value increases
B.Accept the report because specialization implies reliable judgment
C.Require evidence-backed validation before authorizing the action
D.Execute immediately and audit only if a complaint is received
Correct Answer: Require evidence-backed validation before authorizing the action
Explanation:
High-confidence claims are insufficient for irreversible, high-impact actions. Delegation should include evidence requirements and an independent validation gate before authorization.
Incorrect! Try again.
51Several agents must choose values for shared variables subject to cross-agent constraints. They cannot reveal their complete local data. Which approach best supports distributed convergence while preserving locality?
Distributed problem solving
Hard
A.Resolve every constraint using a fixed global value chosen beforehand
B.Exchange constraint-relevant assignments and propagate detected conflicts
C.Send all private data to the agent with the fastest processor
D.Let each agent finalize values without receiving neighbor updates
Correct Answer: Exchange constraint-relevant assignments and propagate detected conflicts
Explanation:
Distributed constraint-solving methods communicate assignments, bounds, or conflicts needed by neighboring agents without necessarily centralizing all private data.
Incorrect! Try again.
52Agents independently solve overlapping subtasks and later discover that their outputs impose contradictory assumptions on a shared variable. Which mechanism would have prevented the most wasted computation?
Distributed problem solving
Hard
A.Random reassignment of all subtasks after contradiction detection
B.Final aggregation after every agent completes its assigned subtask
C.Early propagation of shared constraints and intermediate commitments
D.Removal of shared variables from each agent's local representation
Correct Answer: Early propagation of shared constraints and intermediate commitments
Explanation:
Communicating shared constraints and tentative commitments early exposes incompatible assumptions before agents complete expensive downstream work.
Incorrect! Try again.
53A workflow contains independent research tasks, a synthesis task that depends on all research outputs, and a final verification task. Which orchestration structure best captures these dependencies?
Orchestration patterns
Hard
A.Fan-out research, barrier synchronization, synthesis, then verification
B.Sequential research, verification, fan-out, then independent synthesis
C.Recursive delegation without dependency tracking or completion barriers
D.Peer voting before research, followed by unordered task execution
Correct Answer: Fan-out research, barrier synchronization, synthesis, then verification
Explanation:
Independent research can run in parallel, but synthesis requires a barrier because it depends on all research outputs. Verification should follow the completed synthesis.
Incorrect! Try again.
54A central orchestrator becomes a throughput bottleneck because it routes every message, including routine peer updates. Which redesign reduces this bottleneck while retaining governance over critical actions?
Orchestration patterns
Hard
A.Remove all coordination controls and permit unrestricted side effects
B.Replicate every message to every agent through the same orchestrator
C.Use hierarchical orchestration with peer channels and gated commitments
D.Serialize every subtask so only one agent executes at any time
Correct Answer: Use hierarchical orchestration with peer channels and gated commitments
Explanation:
Peer communication can handle routine coordination, while hierarchical gates retain centralized control over consequential commitments. This reduces routing load without abandoning governance.
Incorrect! Try again.
55Two agents repeatedly acquire resources in opposite orders: agent locks then waits for , while agent locks then waits for . Which coordination rule prevents this deadlock?
Agent coordination mechanisms
Hard
A.Prioritize the agent that most recently acquired its first lock
B.Allow agents to retain acquired locks while requesting more
C.Require all agents to acquire resources in a global order
D.Increase each agent's waiting timeout by the same duration
Correct Answer: Require all agents to acquire resources in a global order
Explanation:
A global lock ordering removes circular wait, one of the necessary conditions for deadlock. Longer timeouts do not eliminate the underlying cycle.
Incorrect! Try again.
56Agents allocate scarce compute through an auction, but one agent exaggerates its private task value to win capacity. Which mechanism property most directly discourages this behavior?
Agent coordination mechanisms
Hard
A.Unlimited capacity granted after the auction has completed
B.Anonymity that hides every bidder's identity permanently
C.Incentive compatibility that makes truthful bidding optimal
D.Random allocation independent of submitted task valuations
Correct Answer: Incentive compatibility that makes truthful bidding optimal
Explanation:
An incentive-compatible mechanism aligns an agent's best strategy with truthful reporting. This directly addresses strategic misrepresentation of private valuations.
Incorrect! Try again.
57In an enterprise purchasing workflow, a requester agent creates an order, an approval agent authorizes it, and a payment agent releases funds. Which control best enforces separation of duties?
Collaborative enterprise workflows
Hard
A.Give every agent shared credentials to simplify audit log correlation
B.Store only the final payment result rather than intermediate decisions
C.Bind each role to distinct credentials and prohibit role self-approval
D.Allow the requester to assume approval rights during peak workloads
Correct Answer: Bind each role to distinct credentials and prohibit role self-approval
Explanation:
Distinct credentials and explicit role constraints prevent one actor from initiating and approving the same transaction. They also produce attributable audit records.
Incorrect! Try again.
58A role-based system allows a research agent to supply evidence and an approval agent to authorize deployment. The research agent embeds an instruction in its evidence asking the approver to bypass policy. What is the strongest architectural defense?
Role-based agents
Hard
A.Treat evidence as untrusted data and enforce approval policy externally
B.Ask the approval agent to obey the most recent instruction it receives
C.Merge both roles so contextual information is never transferred
D.Give the research agent temporary approval privileges for its findings
Correct Answer: Treat evidence as untrusted data and enforce approval policy externally
Explanation:
Cross-role content must be treated as data, not authority. External policy enforcement ensures that embedded instructions cannot expand the sender's privileges.
Incorrect! Try again.
59A system broadcasts every observation to all agents, producing approximately message deliveries per round. Which strategy most directly improves communication scalability when collaboration is sparse?
Scalable AI collaboration strategies
Hard
A.Increase every agent's context window while retaining full broadcast
B.Duplicate each broadcast to improve resilience against message loss
C.Require every agent to acknowledge every observation synchronously
D.Route messages by topic and send them only to relevant agents
Correct Answer: Route messages by topic and send them only to relevant agents
Explanation:
Topic-based selective routing exploits sparse dependencies and avoids all-to-all delivery. It can reduce communication from quadratic growth toward the number of relevant links.
Incorrect! Try again.
60A workflow has two parallel branches with expected durations of 4 and 9 minutes, followed by a 3-minute merge task. Optimizing the 4-minute branch to 1 minute does not change end-to-end latency. Which optimization should be prioritized next?
Workflow optimization for multi-agent systems
Hard
A.Add a barrier before both branches begin their execution
B.Delay the merge task until both branches report twice
C.Reduce the 9-minute branch because it lies on the critical path
D.Further reduce the 4-minute branch because it starts first
Correct Answer: Reduce the 9-minute branch because it lies on the critical path
Explanation:
Parallel completion is bounded by the slower branch, so latency is minutes. Improving the shorter branch does not help until it approaches the critical branch's duration.
Incorrect! Try again.
Did this save you a night before the exam?
LPU Notes is free, and it stays free. Ads cover part of the server bill.
The rest comes out of a student's own pocket: the domain, the storage,
and keeping the site up through the weeks everyone needs it at once.
The payment button didn't load. An ad blocker or a filtered network is the usual reason.
to try again.
Nothing here is ever locked, and nothing unlocks. Chip in only if it was worth it.
What it pays for →