FEFreeExamDumps.in

CCA-F Practice Questions — Page 18

Question 171

Open question ↗

Scenario: Conversational AI Architecture Patterns -

In a long account-support chat that is progressively summarized, the assistant later refers to the customer's plan price as "about thirty dollars" instead of the exact $29.99 and forgets the precise renewal date, leading to an incorrect quote.

What practice preserves these exact values?

  • A.Keep a structured "case facts" block (plan price, renewal date, account ID) included in every prompt, separate from the summarized history.
  • B.Stop summarizing and resend the entire raw transcript on every turn forever.
  • C.Instruct the model to round all prices to whole dollars for consistency.
  • D.Lower the temperature so numeric values stop drifting.

Question 172

Open question ↗

Scenario: Conversational AI Architecture Patterns -

A customer's multi-day support thread spans several sessions. Each new session starts cold, so the assistant re-asks for the order number and prior troubleshooting steps it already collected, frustrating the customer.

Which memory strategy best fixes this across sessions?

  • A.Rely on the model to recall earlier sessions from its training.
  • B.Increase the context window so older sessions stay in memory automatically.
  • C.Tell the customer to repeat the key details at the start of every session.
  • D.Persist key facts and prior steps to structured external storage (a scratchpad/case record) and load them into context at session start.

Question 173

Open question ↗

Scenario: Conversational AI Architecture Patterns -

An assistant must transform user-entered shipping notes into a normalized structure, but the requirement is ambiguous and the team's prose spec keeps being interpreted differently across edge cases.

What most effectively communicates the expected behavior?

  • A.Rewrite the prose specification to be longer and more exhaustive.
  • B.Provide 2–3 concrete input/output examples demonstrating the exact transformation, including a tricky edge case.
  • C.Raise the temperature so the model explores more interpretations.
  • D.Provide a single example and assume the model generalizes perfectly.

Question 174

Open question ↗

Scenario: Conversational AI Architecture Patterns -

During a long investigation chat, verbose tool outputs have filled most of the context window and responses are slowing and degrading. You want to keep the conversation going while accepting a known trade-off.

Which statement about `/compact` is correct?

  • A.`/compact` permanently deletes the conversation and starts a fresh session.
  • B.`/compact` enlarges the model's context window for the rest of the session.
  • C.`/compact` summarizes prior history to reclaim context, but exact numbers, dates, and specifics can be lost in summarization.
  • D.`/compact` stores the full history losslessly and reloads it on demand.

Question 175

Open question ↗

Scenario: Conversational AI Architecture Patterns -

A research assistant synthesizes market figures and reports a single growth rate. Two underlying sources actually disagree: Source A (2023) says 10% and Source B (2024) says 15%. The assistant currently labels this a "contradiction" and picks one value arbitrarily.

What is the correct handling?

  • A.Average the two values to 12.5% and report a single number.
  • B.Always pick the higher value, since a larger figure is more impressive.
  • C.Discard both values because the sources disagree.
  • D.Preserve both values with their publication dates and attribution; the difference may reflect change over time, not a contradiction.

Question 176

Open question ↗

Scenario: Conversational AI Architecture Patterns -

A multi-turn assistant's context is filling quickly because `get_account` returns 50+ fields per call but each turn needs only 4 of them. Responses are starting to slow and occasionally lose track of earlier details.

What is an effective mitigation?

  • A.Trim each tool result down to the few relevant fields before it enters context (e.g., via a `PostToolUse` hook).
  • B.Tell the model to ignore the fields it doesn't need.
  • C.Switch to a larger-context model and keep returning all 50 fields.
  • D.Call the tool once and cache all 50 fields for the rest of the chat.

Question 177

Open question ↗

Scenario: Conversational AI Architecture Patterns -

A conversational form-filler sometimes answers ambiguous questions in free text instead of emitting the structured `submit_form` payload the backend expects, causing intermittent integration failures.

How do you guarantee a structured tool call on that turn?

  • A.Add a system-prompt line asking the model to "always return the form payload."
  • B.Set `tool_choice: "any"` so the model must call a tool rather than reply with free text.
  • C.Lower the temperature so the model stops writing prose.
  • D.Increase `max_tokens` so the structured output has room to complete.

Question 178

Open question ↗

Scenario: Conversational AI Architecture Patterns -

Mid-conversation a user types: "This isn't working — just get me a human supervisor." The assistant continues asking diagnostic questions and runs two more tool calls before considering escalation.

What is the correct behavior?

  • A.Keep troubleshooting; only escalate if the assistant exhausts its own options.
  • B.Run a sentiment classifier to verify the user is upset before escalating.
  • C.Ask the user to confirm the request two more times to avoid a false escalation.
  • D.Escalate immediately — an explicit request for a human is an immediate-escalation trigger.

Question 179

Open question ↗

Scenario: Conversational AI Architecture Patterns -

A research assistant states in its answer: "The global AI tutoring market is about $4.1B." It gives no source or date, so reviewers cannot verify it and cannot reconcile it against a different figure another source provided.

How should claims be represented to preserve provenance?

  • A.Summarize all claims into one fluent paragraph for readability.
  • B.Keep only the highest figure, since it is probably the most recent.
  • C.Attach a structured claim→source mapping: source name, URL, publication date, and confidence for each claim.
  • D.Drop any claim that has more than one possible source.

Question 180

Open question ↗

Scenario: Conversational AI Architecture Patterns -

A banking assistant asks "Which account do you mean?" after a vague request, and the `find_accounts` tool returns two checking accounts that both match the customer's name. The assistant is about to act on the first one returned.

What is the correct handling of the multiple matches?

  • A.Act on the first matching account, since it was returned first.
  • B.Merge the two accounts' balances and present a combined view.
  • C.Pick whichever account had the most recent activity as a tiebreaker.
  • D.Ask the customer for an additional identifier (last four digits or account nickname) to disambiguate before acting.