Question 220
A validator rejects an output because a field expected to be a number contains a currency symbol. What contract clarification is most relevant?
- Delete the amount field and assume downstream software will guess it
- Define numeric values separately from currency codes or display formatting
- Change the entire record to arbitrary text to accept the symbol
- Treat invalid JSON types as automatically harmless
Explanation
Correct Answer
B:Machine-readable amounts should use a consistent representation. Presentation symbols are different from the value's type.
Incorrect Answers
A:An omitted required amount cannot support totals or payment checks. The value must be explicit.
C:Relaxing all typing loses validation benefits. Fix the offending field representation instead.
D:Type mismatches can produce runtime errors or incorrect calculations. The schema should be honored.
Question 221
A system extracts purchase orders into JSON and then releases payments. Which THREE independent assurances are needed before a high-value payment can be authorized? Choose THREE.
- Assume the presence of valid JSON proves every amount is authentic
- Cross-check material values and totals against the source and business rules
- Allow the model to invent any missing mandatory identifier
- Apply a separate authorization policy before the financial side effect
- Treat the model's positive tone as an authorization token
- Validate the record against a defined structured schema
Explanation
Correct Answers
B:A schema-valid amount can still be wrong. Source comparison and deterministic arithmetic help catch semantic errors.
D:Even a correctly parsed invoice may not be approved for payment. Execution privileges must be checked outside model generation.
F:Typed fields and requiredness reduce malformed outputs and parsing ambiguity. This addresses representation, not truth by itself.
Incorrect Answers
A:A parser validates shape, not financial truth. It cannot replace source evidence and business checks.
C:Synthetic values can connect a payment to the wrong record. Missing information requires correction or review.
E:A natural-language opinion cannot grant financial permissions. Approval depends on trusted policy and identity evidence.
Question 222
A workflow needs two alternative shapes: success with data or failure with an error category. How should its output contract be designed?
- Omit status entirely and rely on whether the text sounds positive
- Use an explicit discriminated status with defined fields for each outcome
- Return success-shaped placeholder data on all failures
- Use arbitrary key names chosen on each request
Explanation
Correct Answer
B:Separating successful records from failures prevents consumers from treating errors as ordinary data. The validator can enforce each variant.
Incorrect Answers
A:Prose is not a reliable parser contract. A clear outcome field enables deterministic handling.
C:Dummy values can silently enter business processes. Error outcomes need a distinct representation.
D:Changing structures defeats stable integrations. Both branches should be documented and validated.
Question 223
A developer wants every generated JSON object to include a source reference for each extracted field. What schema refinement is most useful?
- Treat the presence of any URL as proof that all fields are sourced
- Store all sources in an undocumented free-text comment
- Ask the model to remember the source mentally without emitting references
- Store values with provenance spans or source IDs
Explanation
Correct Answer
D:Field-level provenance supports auditing and correction of disputed extractions. The contract should specify which evidence format the validator accepts.
Incorrect Answers
A:One generic link does not validate every value. References must actually support the corresponding fields.
B:Unstructured comments make it hard to pair each field with support. Typed provenance enables systematic checks.
C:Invisible recollection cannot be audited by a downstream system. Evidence should be part of the output.
Question 224
A downstream system parses JSON keys case-sensitively, but the model occasionally changes customer_id to CustomerID. Which design choice helps?
- Define canonical field names in a strict output schema
- Remove all identifiers from the output to avoid the mismatch
- Tell the parser to accept every vaguely similar key without a mapping
- Randomize field casing in examples to show flexibility
Explanation
Correct Answer
A:Schema-driven generation and parsing reduce key drift. Consumers should reject unexpected names rather than silently misroute data.
Incorrect Answers
B:Identifiers are essential to the business record. Fix the naming discipline rather than dropping data.
C:Heuristic matching can conflate different identifiers. A deliberate compatibility map is safer if aliases are required.
D:Inconsistent examples make the requested stability harder. The contract requires exact names.
Question 225
A team needs to collect multiple independent contacts from a form and preserve blanks explicitly. How should the structured representation handle missing data?
- Require every record to contain guessed nonempty values
- Concatenate all contacts into an unseparated paragraph
- Represent contacts as an array with defined optional fields
- Invent contact information for every absent field
Explanation
Correct Answer
C:The schema should state how optional values are represented and distinguish missing entries from empty strings when that matters.
Incorrect Answers
A:A nonempty constraint without source evidence encourages hallucination. Business validation should handle incompleteness honestly.
B:A paragraph loses contact boundaries and field meanings. The downstream process expects structured entries.
D:Synthetic values misrepresent the source and can trigger wrong downstream actions. Missingness must remain explicit.
Question 226
A schema-valid extraction reports invoice total 900, but the document states 90. What validation step catches the problem?
- Check extracted values against sources and calculations
- Ask the model to produce more fluent JSON labels
- Remove amount checks to prevent validation failures
- Assume the value is correct because it parsed as a number
Explanation
Correct Answer
A:Structural validity does not guarantee semantic correctness. Source-grounded checks detect misplaced digits or inconsistent totals.
Incorrect Answers
B:Label style is unrelated to matching the invoice value. Evidence validation is required.
C:Suppressing checks hides real extraction defects. The pipeline should repair or escalate them.
D:Numeric type checks only confirm representation. The amount can still be factually wrong.
Question 227
A structured output fails a JSON schema because a required field is missing. What retry strategy is most effective?
- Insert a fabricated value to force schema compliance
- Give field-specific feedback and retry a bounded number of times
- Declare success because other fields were present
- Repeat the original prompt unchanged indefinitely
Explanation
Correct Answer
B:Concrete errors tell the model what to repair. Retry limits prevent infinite loops and trigger escalation when data is unavailable.
Incorrect Answers
A:A valid shape with invented data is not a correct extraction. Missing evidence must remain visible.
C:The contract explicitly requires the missing field. Partial completion is not a valid full record.
D:Unmodified retries may reproduce the same omission. The model needs targeted feedback or missing-source clarification.
Question 228
A document lacks a mandatory tax ID, but the model keeps guessing one after each failed validation. What should the workflow do?
- Increase retry count until a plausible ID appears
- Report missing source data for correction or review
- Relax the type to accept any fictional identifier
- Suppress validation messages from audit logs
Explanation
Correct Answer
B:Validation cannot conjure absent evidence. The correct outcome is incomplete data with an explicit blocker, not invented identifiers.
Incorrect Answers
A:More generations do not create authoritative source evidence. Repeated guessing increases hallucination risk.
C:Changing representation does not solve the missing fact. The business requirement remains unmet.
D:Hiding failures prevents diagnosis and allows bad data downstream. Errors should be retained safely.
Question 229
A generated order has quantities and unit prices but its total does not equal the sum of line items. Which check is required?
- Discard line items and keep only the total
- Ask for more descriptive product names instead
- Trust the model's total because the JSON object is complete
- Recalculate totals in code and compare to model output
Explanation
Correct Answer
D:Arithmetic validation can identify inconsistencies independent of model confidence. Mismatches should trigger correction or review.
Incorrect Answers
A:Removing evidence prevents reconciliation and may preserve the wrong value. The record needs its supporting components.
B:Longer names do not fix arithmetic. The check should target the computed relationship.
C:Completeness of fields does not establish numerical consistency. A valid schema can carry invalid calculations.
Question 230
A parser repeatedly fails on the same malformed document after two informed retries. What should the workflow do next?
- Return a controlled validation failure for human correction
- Publish a fabricated complete record to meet the deadline
- Continue retries without an upper limit
- Reset all error counters after each attempt
Explanation
Correct Answer
A:Bounded retries avoid runaway cost. Persistent failures may require better source quality or human judgment.
Incorrect Answers
B:False data creates downstream business risk. Honest failure is preferable to invented success.
C:Repeated model calls can loop indefinitely on irreparable input. The system needs a stop condition.
D:Counter reset defeats the retry policy and obscures repeated failures. Persistent state is required.