Amazon AWS Certified Generative AI Developer - Professional AIP-C01 Practice Test 49

Question 480

A chatbot passes small JSON cases but returns unfinished objects for long requests. Runtime traces show a max-token stop reason at a low configured output-token limit. Which diagnostic can confirm the mechanism and guide a safe retest?

Parse an incomplete JSON prefix as a successful result because generation ended normally at the token limit.
Raise only the API Gateway timeout while maxTokens remains unchanged.
Treat closing braces guessed by the application as evidence that the missing output fields were actually generated.
Correlate stop reason and token usage with the configured/model-supported output limit, then retest with sufficient bounded output budget and schema validation.

Explanation

Correct Answer

When the response stops at max_tokens, the immediate failure is output truncation, not necessarily a schema-model bug. Inspect token usage and allowed limits before evaluating a corrected full response.

Incorrect Answers

A length-limited stop can leave required fields missing.

Timeout settings cannot extend the model's configured output token cap.

Auto-completing syntax does not validate business semantics.

Question 481

A document answer omits latter pages. Which diagnostic distinguishes the likely service failure?

Assume the model hallucinated when the PDF parser never extracted the missing pages.
Increase top-k without checking whether the omitted pages were indexed at all.
Diagnose context overflow, parser truncation and retrieved chunk budget.
Replace the document source with model memory of similar reports.

Explanation

Correct Answer

Trace parser page coverage, retrieved chunks and model context/token limits to locate omitted latter pages before adjusting the relevant stage.

Incorrect Answers

Generation cannot cite evidence absent from ingestion.

Retrieval tuning cannot recover missing source text.

Pretraining does not establish the missing exact pages.

Question 482

A Knowledge Base’s indexed document embeddings still use version V1, but a recently deployed query encoder uses model version V2; vectors have the same dimensionality, yet retrieval quality collapses. Before rebuilding the entire index, which diagnostic confirms the mismatch?

Keep old corpus vectors but change only the query embedding model and assume same vector length means compatibility.
Increase the reranker candidate count while initial nearest neighbors are based on incompatible embeddings.
Change the cache TTL without checking which model generated stored vectors.
Compare the active query-embedding model/version with the version used for stored corpus embeddings, then validate a compatible rollback or complete reindexing plan.

Explanation

Correct Answer

Equal vector dimensionality does not prove semantic embedding compatibility. Inspect the model version used at indexing versus query time before deciding whether to restore V1 queries or rebuild the corpus with V2.

Incorrect Answers

Equal dimensionality does not prove semantic-space equivalence.

Reranking cannot recover missing gold hits excluded by invalid candidate retrieval.

Freshness settings do not correct embedding-space mismatch.

Question 483

Bedrock Knowledge Bases returns highly similar passages from the wrong insurance product line although the client supplied a business-domain metadata filter. The filter field does not match the indexed metadata schema. What retrieval configuration should the engineer inspect?

Increase model output tokens to encourage it to avoid irrelevant retrieved documents.
Check that the metadata filter uses the indexed product-domain field and allowed values, then compare filtered retrieval and hybrid ranking behavior.
Treat high similarity score as proof the hit belongs to the intended policy domain.
Disable metadata filters altogether because they might remove relevant passages.

Explanation

Correct Answer

A high semantic score does not enforce business scope. A mismatched or unapplied metadata field can admit passages from another domain; verify filter semantics before blaming generation.

Incorrect Answers

Output capacity cannot correct source selection at retrieval time.

Semantic relatedness does not establish permission or department scope.

Removing domain boundaries can worsen cross-domain contamination.

Question 484

A user receives an outdated answer even though KB was synced. Which diagnostic distinguishes the likely service failure?

Invalidate stale application caches and verify indexing/source effective dates.
Run the same cached answer formatter repeatedly without fetching current source evidence.
Change FM temperature while keeping the stale result cache untouched.
Delete only the S3 source version reference but retain cached generated answers indefinitely.

Explanation

Correct Answer

After KB source synchronization, invalidate application caches and verify index/source effective dates; a stale cached answer can survive a fresh index.

Incorrect Answers

The app can continue returning obsolete text even when indexing succeeded.

Sampling does not update cached policy text.

Removing a reference does not remove old cached content.

Question 485

After a Knowledge Base changed to hierarchical chunking, an operations dashboard shows three returned passages instead of eight. The incident commander proposes re-ingesting all S3 source documents. Which diagnostic should be completed before treating this as an ingestion failure?

Reingest the entire corpus automatically whenever final hits are fewer than requested.
Assume numberOfResults is the exact final number of distinct parent passages.
Disable parent linkage and assume relevance will stay unchanged.
Trace child-to-parent replacement for the affected query and compare retrieval results with source-ingestion status before approving a full re-ingestion.

Explanation

Correct Answer

A lower count of final parent passages can be normal replacement of multiple child hits. Distinguish this retrieval behavior from ingestion loss using the parent mapping and ingestion/source inventory before destructive reprocessing.

Incorrect Answers

Parent replacement can produce this behavior without missing documents.

It limits child retrieval; returned parent count may differ.

The parent context can be essential for grounded answers.

Question 486

A reranker improves precision but doubles response latency. What specific check can confirm the actual failure mechanism?

Remove all source-quality checks solely because reranking adds latency.
Rerank thousands of irrelevant candidates with an unlimited context budget.
Profile candidate count and reranker invocation without removing all retrieval checks.
Increase the answer model's temperature to offset reranking work.

Explanation

Correct Answer

Profile reranker input candidate count, supported model invocation and p95 tradeoff while retaining measured precision and acceptable answer context.

Incorrect Answers

Faster but ungrounded output fails the goal.

Overlarge candidate sets amplify reranker cost/latency.

Sampling parameters do not remove reranker execution time.

Question 487

Immediately after a tool deployment, an AgentCore Gateway-backed agent gets parameter-schema validation errors. The deployed target requires a new field, but the agent still sends the previous request shape. Select TWO checks that identify both the deployed contract and the invalid client request.

Compare deployed function/tool schema version.
Disable schema validation so the model can pass arbitrary extra tool parameters.
Give the tool execution role broader IAM permissions to fix a missing required field.
Assume a successful model-generated toolUse block proves the new function schema is compatible.
Validate actual arguments against current required/optional fields.

Explanation

Correct Answer

Compare the deployed tool/function version and its current schema contract.

Validate the exact emitted parameters against required fields, type rules and allowed optional arguments.

Incorrect Answers

Unchecked arguments can break or abuse the backend contract.

Authorization cannot repair request shape.

Tool intent may still fail validation when executed.

Question 488

A Step Functions workflow retries a payment write twice. Which troubleshooting step should be performed first?

Diagnose idempotency and Catch/Retry semantics to avoid duplicate transactions.
Add a fresh payment ID for every retry after a timeout.
Disable transaction logging to make the state-machine retry graph look simpler.
Treat Retry configuration as automatic rollback of already committed payment side effects.

Explanation

Correct Answer

Trace Step Functions Catch/Retry attempts and the payment idempotency key to determine why a write was committed twice; reconcile actual ledger state.

Incorrect Answers

A prior attempt may have committed, causing duplicate charges.

Removing evidence prevents financial reconciliation.

Workflow retry does not reverse a prior backend transaction.

Question 489

An app fails intermittently at high concurrent load. What specific check can confirm the actual failure mechanism?

Change the prompt tone when the service returns ThrottlingException at high concurrency.
Distinguish throttling, quota and backend dependency latency from model output quality.
Measure only average response correctness without concurrent-load error codes.
Add unlimited retries that compete with live user traffic.

Explanation

Correct Answer

Inspect throttling codes, regional quotas, actual token demand and backend/tool latency under load before attributing failures to model quality.

Incorrect Answers

The error occurs at resource admission, not in generated reasoning.

Quality metrics cannot diagnose capacity/quota saturation.

A retry storm can worsen concurrency failures.

Question 490

A prompt variant generates different field labels. Which diagnostic distinguishes the likely service failure?

Assume every versioned prompt generates identical key names because its purpose is unchanged.
Compare template/model versions and validate structured schema across releases.
Accept any JSON field name that sounds similar and silently write it into the old schema.
Change only the UI label without testing the API's actual output fields.

Explanation

Correct Answer

Compare saved prompt versions, model/request settings and rendered field mappings; validate structured output contract across the new variant.

Incorrect Answers

Template changes can alter field labels and response shape.

Ambiguous field mapping can corrupt business data.

A display change does not repair a machine-readable contract regression.

Pass Your Amazon AWS Certified Generative AI Developer - Professional AIP-C01 Exam Easy!

Question 480

Explanation

Correct Answer

Incorrect Answers

Question 481

Explanation

Correct Answer

Incorrect Answers

Question 482

Explanation

Correct Answer

Incorrect Answers

Question 483

Explanation

Correct Answer

Incorrect Answers

Question 484

Explanation

Correct Answer

Incorrect Answers

Question 485

Explanation

Correct Answer

Incorrect Answers

Question 486

Explanation

Correct Answer

Incorrect Answers

Question 487

Explanation

Correct Answer

Incorrect Answers

Question 488

Explanation

Correct Answer

Incorrect Answers

Question 489

Explanation

Correct Answer

Incorrect Answers

Question 490

Explanation

Correct Answer

Incorrect Answers

AWS Certified Generative AI Developer - Professional AIP-C01 Practice Tests