When analyzing complex legal cases that cite multiple precedents, the document analysis subagent processes each sequentially. A landmark case citing 12 precedents takes over 3 minutes to analyze completely. What's the most effective way to reduce this latency while preserving the coordinator's ability to monitor and debug the system?
Answer(s): D
A: Enable the document analysis subagent to spawn its own specialized subagents dynamically when it encounters cases with many citations. Incorrect.This decentralizes orchestration and makes the system harder to monitor and debug. The coordinator loses visibility into dynamically spawned agents.B: Implement a message queue where precedent analysis tasks are processed asynchronously by a pool of worker agents. Incorrect.While this improves scalability, it introduces infrastructure complexity and reduces transparency for debugging at the coordinator level.C: Create a recursive agent hierarchy where analysis agents subdivide work among child agents until reaching single-precedent granularity. Incorrect.This further complicates the architecture and makes tracing execution paths difficult, reducing observability and control.D: Have the coordinator spawn parallel document analysis subagents, each handling a subset of precedents, then aggregate results before synthesis. Correct.This enables parallel processing to reduce latency while keeping orchestration centralized. The coordinator retains full visibility, making monitoring and debugging easier.
Introduction monitoring shows the research phase takes longer than expected. Analysis reveals the coordinator invokes the web search subagent, waits for its response, then invokes the document analysis subagent and waits again. These tasks are independent - neither requires the other's output. How should you modify the system to run these subagents concurrently?
A: Switch both subagents to use a Haiku tier model instead to reduce their individual execution time. Incorrect.This may reduce latency per task, but it does not address the core issue of sequential execution vs. parallelism.B: Create an async orchestration layer outside the agent that spawns parallel threads, each running a separate coordinator subagent pair, then aggregates results. Incorrect.This overcomplicates the architecture and duplicates coordinators unnecessarily instead of fixing concurrency within the existing flow.C: Add detailed instructions to the coordinator's system prompt explaining the performance benefits of parallel execution and requesting it invoke both subagents at the same time. Incorrect.Instructions alone are not reliable for enforcing concurrency. Execution behavior depends on how tool calls are structured, not just prompt wording.D: Structure the coordinator to emit both Task tool calls (for web search and document analysis) in a single response message rather than across separate conversation turns. Correct.Issuing both tool calls in one response enables true parallel execution , since the system can run them concurrently instead of waiting for one to finish before starting the other.
Production reviews reveal inconsistent handling of uncertainty in final reports. Sometimes conflicting subagent findings are synthesized into a single confident statement (losing other times reports over hedge with excessive qualifications (becoming unhelpful). When the web search agent returns "Industry analysts estimate $50B market size (methodolo the document analysis agent returns "peer-reviewed study estimates $358 (1578, 95% CI)," the coordinator either picks one arbitrarily or produces vague statements like "the ma 6358-6508 depending on factors." What systematic approach best addresses this?
Answer(s): C
A: Configure subagents to only report findings meeting a high confidence threshold, filtering uncertain information before it reaches the coordinator. Incorrect.This suppresses potentially valuable but uncertain insights and introduces bias by hiding ambiguity rather than managing it.B: Add a verification subagent that cross-references findings across sources, only passing claims to synthesis that are corroborated by at least two independent sources. Incorrect.While useful for validation, this approach still filters out uncertainty instead of representing it, and may discard novel or emerging insights.C: Instruct the synthesis agent to structure reports with explicit sections distinguishing well-established findings from contested ones, preserving original source characterization and methodological context. Correct.This directly addresses inconsistent handling of uncertainty by making it explicit and structured , allowing users to understand both consensus and disagreement without losing context.D: Implement a confidence calibration layer that normalizes subagent uncertainty expressions to standardized probability scores (0.0–1.0), then weight-average findings by their calculated reliability scores to produce a statistically grounded synthesis. Incorrect.This introduces artificial precision and may oversimplify complex, qualitative uncertainty, potentially misleading users.
In production, you observe that simple fact-checking queries (e.g., "What year was the Paris Climate Agreement signed?") traverse all four subagents sequentially, consuming 40+ seconds. While this might be acceptable for complex comparative research benefits from the full pipeline. Your query distribution is diverse and evolving as users discover new applications. What's the most effective approach to optimize for varying query complexity?
A: Implement pattern-based routing that categorizes queries by structure (single-fact vs. comparative vs. analytical) and maps each category to a predefined subagent combination. Incorrect.This is rigid and brittle. As query patterns evolve, maintaining rules becomes difficult and coverage gaps are likely.B: Train a query complexity classifier on labeled historical data to predict optimal subagent combinations, retraining periodically as query patterns evolve. Incorrect.While adaptive, this introduces model maintenance overhead , requires labeled data, and may lag behind new or rare query types.C: Have the coordinator analyze each query and dynamically decide which subagents to invoke based on its assessment of query requirements. Correct.This provides flexible, real-time routing without rigid rules or heavy ML infrastructure. The coordinator can tailor execution paths to query complexity efficiently.D: Create a fast-path for factual questions that bypasses subagents entirely, routing all other queries through the complete pipeline to ensure research thoroughness. Incorrect.This is overly simplistic and risks misclassification, reducing accuracy or missing nuance for queries that appear simple but require deeper analysis.
A user is expanding the research system beyond its single web search agent by adding specialized data sources. They add a financial API agent that returns structured JSON with margins, and growth rates; a news monitoring agent that returns prose summaries of recent developments; and a patent analysis agent that returns structured lists of technology synthesis agent combines these into executive briefings. Currently, it converts everything to bullet points, causing financial comparisons to lose tabular clarity and news summari narrative flow. What change would most improve briefing quality?
Answer(s): A
A: Update the synthesis agent to render each content type appropriately—financial data as tables, news as prose, and technical lists as structured points. Correct.This preserves the natural structure and strengths of each data type , improving clarity, readability, and usefulness of the final briefing.B: Add a format conversion layer between subagents and synthesis that transforms all outputs to a common intermediate representation (such as Markdown) to facilitate more flexible rendering. Incorrect.While helpful for consistency, this does not guarantee appropriate presentation of different content types and may still lead to generic formatting.C: Standardize all subagent outputs to JSON with fields for every data type to ensure programmatic consistency across the pipeline. Incorrect.This improves structure but shifts complexity to the synthesis stage and does not inherently improve human-readable output quality .D: Standardize all subagent outputs to prose summaries with a uniform character to maintain a consistent executive voice regardless of the source material. Incorrect.This sacrifices important structure (like tables and lists), reducing clarity and effectiveness for data-heavy content.
Post your Comments and Discuss Anthropic CCAR-F exam prep with other Community members: