Agent Topologies & Orchestration•20% of Exam Score
Multi-Agent Research System Design
High-throughput autonomous research requires coordinating specialized models without suffering from context window explosion, hallucination cascading, or communication deadlocks.
Exam Syllabus Deep Dive
Core Architectural Principles Tested
Topic 1
Coordinator-Worker Topologies
A lead coordinator decomposes complex queries and delegates subtasks to specialized worker agents.
Architectural Best Practice:Keep the coordinator prompt focused purely on task planning, routing, and synthesis. Do not burden it with raw research data.
Topic 2
Fan-out / Fan-in Parallel Tool Execution
Executing research searches sequentially introduces unacceptable latency in deep investigations.
Architectural Best Practice:Emit parallel tool calls across subagents and synthesize structured outputs in a synchronized aggregation step.
Topic 3
Context Window Compaction & Truncation
Subagent search transcripts quickly consume hundreds of thousands of tokens if returned raw.
Architectural Best Practice:Enforce worker agents to summarize their findings into concise, citation-backed digests before returning to the lead coordinator.
High-Yield Exam Rules to Remember:
- Coordinator models must never receive raw unstructured tool outputs from sub-agents.
- Set hard recursion depth limits on sub-agent spawning to avoid exponential call tree runaway.
- Use structured JSON outputs for inter-agent communication to ensure deterministic parsing.
Interactive Practice
Sample Questions from this Domain (18 Total)
Sample Question #1 (Q#7)Multi-agent Research System
Situation Context:The system researches “AI impact on creative industries,” but reports cover only visual art. The coordinator decomposed the topic into: “AI in digital art,” “AI in graphic design,” “AI in photography.”
What’s the cause?
AThe synthesis agent does not detect gaps
BThe coordinator decomposed the task too narrowly
CThe web search agent does not search thoroughly enough
DThe document analysis agent filters out non-visual sources
Explanation & Architectural Rationale:The logs show the coordinator decomposed “creative industries” only into visual subtopics, completely missing music, literature, and film. Subagents executed correctly—the issue is what they were assigned.
Sample Question #2 (Q#8)Multi-agent Research System
Situation Context:A web-search subagent times out while researching a complex topic. You need to design how error information is passed back to the coordinator.
Which error propagation approach best enables intelligent recovery?
AReturn structured error context to the coordinator: failure type, query, partial results, and alternatives
BImplement automatic retries with exponential backoff inside the subagent, then return a generic “search unavailable” status
CCatch the timeout inside the subagent and return an empty result set marked as success
DPropagate the timeout exception to a top-level handler that terminates the whole workflow
Explanation & Architectural Rationale:Structured error context gives the coordinator what it needs to decide whether to retry with a modified query, try an alternative approach, or continue with partial results. B hides context behind a generic status. C masks failure as success. D aborts the entire workflow unnecessarily.
Sample Question #3 (Q#9)Multi-agent Research System
Situation Context:The synthesis agent often needs to verify specific claims while merging results. Currently, when verification is needed, the synthesis agent hands control back to the coordinator, which calls the web-search agent and then re-runs synthesis with the new results. This adds 2–3 extra round trips per task and increases latency by 40%. Your assessment shows that 85% of these checks are simple fact checks (dates, names, statistics), while 15% require deeper investigation.
How do you reduce overhead while maintaining reliability?
AGive the synthesis agent a limited `verify_fact` tool for simple checks, and continue routing complex verification through the coordinator
BAccumulate all verification needs into a batch and return them to the coordinator at the end
CGive the synthesis agent full access to all web-search tools
DProactively cache additional context around each source
Explanation & Architectural Rationale:This applies the principle of least privilege: the synthesis agent gets exactly what it needs for the 85% common case (simple fact checks) while preserving the coordinator-mediated path for complex investigations. B introduces blocking dependencies (later synthesis steps may depend on earlier verified facts). C breaks separation of responsibilities. D relies on speculative caching that cannot reliably predict needs.