Snowflake SnowPro Specialty Gen AI GES-C01 Practice Test 18
Question 170
At an international retail distributor, a Snowflake engineering team is undertaking a multilingual support rollout. A downstream API rejects model replies because some fields are numeric in one record and strings in another. The proposed fix is still hypothetical; regional staff use different documentation versions, and the regulator expects audit trails for significant decisions. Which TWO checks provide the most decisive evidence before approving a change?
- Compare raw generated payload to JSON schema
- Compare train and test similarity
- Inspect memory peaks and batch dimensions
- Inspect invalid records across model versions.
- Inspect deployed version identifiers
Explanation
Correct Answers
A:This observation directly tests the case mechanism: Descriptive prompts alone do not guarantee stable types or required fields. In the described setting, it helps determine whether configure structured output expectations and validate each returned object against an application schema. is justified. These author-intended interpretations require independent technical review before operational use.
D:This complementary observation tests a different part of the same hypothesis and should be read together with the first check before authorizing change. Expected success would be: every accepted response has required fields and correct data types. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
B:Checking this might help diagnose fine-tune evaluation, where a custom model improves scores on examples that are nearly identical to its training records. It is not one of the two most direct checks for the present case because the leading mechanism is descriptive prompts alone do not guarantee stable types or required fields. These author-intended interpretations require independent technical review before operational use.
C:Checking this might help diagnose snowpark service resource sizing, where a containerized embedding model works on small inputs but evicts under parallel batch requests. It is not one of the two most direct checks for the present case because the leading mechanism is descriptive prompts alone do not guarantee stable types or required fields. These author-intended interpretations require independent technical review before operational use.
E:Checking this might help diagnose model registry rollback, where a production release degrades document extraction and the team cannot tell which model artifact is active. It is not one of the two most direct checks for the present case because the leading mechanism is descriptive prompts alone do not guarantee stable types or required fields. These author-intended interpretations require independent technical review before operational use.
Question 171
At a specialty chemicals wholesaler, a Snowflake engineering team is undertaking a catalog embedding prototype. A downstream API rejects model replies because some fields are numeric in one record and strings in another. Diagnosis is disputed because product names include rare identifiers; retrieval relevance is more important than creative text. Which underlying failure mechanism BEST explains the observed limitation?
- Descriptive prompts alone do not guarantee stable types or required fields.
- Sampling controls alone cannot guarantee correct numbers without programmatic validation.
- A model invocation cannot safely infer mismatched feature names or types.
- Unversioned generated labels cannot be reproduced or adjudicated.
Explanation
Correct Answer
A:This is the causal hypothesis consistent with the observed limitation: a downstream api rejects model replies because some fields are numeric in one record and strings in another. Confirm it through both compare raw generated payload to json schema and inspect invalid records across model versions. before changing the service. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
B:That mechanism could explain a different case about response temperature and repeatability, but does not most directly account for this specific symptom: a downstream api rejects model replies because some fields are numeric in one record and strings in another. Here, the evidence to collect is compare raw generated payload to json schema These author-intended interpretations require independent technical review before operational use.
C:That mechanism could explain a different case about model registry signature mismatch, but does not most directly account for this specific symptom: a downstream api rejects model replies because some fields are numeric in one record and strings in another. Here, the evidence to collect is compare raw generated payload to json schema These author-intended interpretations require independent technical review before operational use.
D:That mechanism could explain a different case about prompt result auditing, but does not most directly account for this specific symptom: a downstream api rejects model replies because some fields are numeric in one record and strings in another. Here, the evidence to collect is compare raw generated payload to json schema These author-intended interpretations require independent technical review before operational use.
Question 172
At a fast-growing education platform, a Snowflake engineering team is undertaking a learner feedback intelligence program. A downstream API rejects model replies because some fields are numeric in one record and strings in another. The change was trialed on a small dataset; team leaders have strict response-time targets, while training cohorts use several languages. Which acceptance condition BEST demonstrates that the intervention worked?
- Every accepted response has required fields and correct data types.
- The team reports sentiment separately from conversion predictions.
- End-to-end requests meet latency after retrieval optimizations.
- Repeated reports agree on computed amounts and expose supported wording.
Explanation
Correct Answer
A:This is the observable acceptance condition that would show the intended fix addressed descriptive prompts alone do not guarantee stable types or required fields. It is stronger than activity counts because it tests the user-facing effect. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
B:This could demonstrate success for sentiment numeric interpretation, but it would not establish resolution of the current issue: a downstream api rejects model replies because some fields are numeric in one record and strings in another. Validate every accepted response has required fields and correct data types. instead. These author-intended interpretations require independent technical review before operational use.
C:This could demonstrate success for cortex search latency, but it would not establish resolution of the current issue: a downstream api rejects model replies because some fields are numeric in one record and strings in another. Validate every accepted response has required fields and correct data types. instead. These author-intended interpretations require independent technical review before operational use.
D:This could demonstrate success for response temperature and repeatability, but it would not establish resolution of the current issue: a downstream api rejects model replies because some fields are numeric in one record and strings in another. Validate every accepted response has required fields and correct data types. instead. These author-intended interpretations require independent technical review before operational use.
Question 173
At a charitable grant foundation, a Snowflake engineering team is undertaking a grant document review workflow. A downstream API rejects model replies because some fields are numeric in one record and strings in another. An auditor will inspect the control design; compliance staff need audit evidence, and the archive includes superseded contract amendments. Which ongoing control MOST directly reduces the original risk?
- Reject invalid payloads before writing business records.
- Require versioned deploy manifests and approval logs.
- Add preflight token budgets and truncation alarms.
- Version synonyms with business-definition review.
Explanation
Correct Answer
A:This retains a specific safeguard against recurrence of descriptive prompts alone do not guarantee stable types or required fields. It complements the immediate intervention, configure structured output expectations and validate each returned object against an application schema., and allows the team to verify the intended result. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
B:This is a sensible safeguard for model registry rollback, but it does not target the concrete exposure in this scenario. The current failure is descriptive prompts alone do not guarantee stable types or required fields. These author-intended interpretations require independent technical review before operational use.
C:This is a sensible safeguard for llm prompt token counting, but it does not target the concrete exposure in this scenario. The current failure is descriptive prompts alone do not guarantee stable types or required fields. These author-intended interpretations require independent technical review before operational use.
D:This is a sensible safeguard for cortex analyst schema ambiguity, but it does not target the concrete exposure in this scenario. The current failure is descriptive prompts alone do not guarantee stable types or required fields. These author-intended interpretations require independent technical review before operational use.
Question 174
At a national furniture retailer, a Snowflake engineering team is undertaking an automated customer feedback review. A downstream API rejects model replies because some fields are numeric in one record and strings in another. Budget and operational risk are both material; promotion teams use rapidly changing product catalogs; managers want segment-specific reporting. Which decision MOST appropriately balances the requested outcome with production constraints?
- Configure structured output expectations and validate each returned object against an application schema. Then reject invalid payloads before writing business records.
- Explain the scoring scale and validate its relationship to the actual business label. Postpone the critical evidence checks until after a broad release.
- Benchmark memory and concurrency on a suitable Snowpark Container Services compute pool. Postpone the critical evidence checks until after a broad release.
- Store approved prompt, model version, source identifiers, and evaluation outcome with each label. Postpone the critical evidence checks until after a broad release.
Explanation
Correct Answer
A:This combines a corrective action with a bounded production safeguard, addressing a downstream api rejects model replies because some fields are numeric in one record and strings in another. Prove the result by checking that every accepted response has required fields and correct data types. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
B:This proposal begins with an intervention for sentiment numeric interpretation and postpones evidence that is essential for the current case. Before rollout, the team should configure structured output expectations and validate each returned object against an application schema. These author-intended interpretations require independent technical review before operational use.
C:This proposal begins with an intervention for snowpark service resource sizing and postpones evidence that is essential for the current case. Before rollout, the team should configure structured output expectations and validate each returned object against an application schema. These author-intended interpretations require independent technical review before operational use.
D:This proposal begins with an intervention for prompt result auditing and postpones evidence that is essential for the current case. Before rollout, the team should configure structured output expectations and validate each returned object against an application schema. These author-intended interpretations require independent technical review before operational use.
Question 175
At a global food wholesaler, a Snowflake engineering team is undertaking a vendor invoice reconciliation. A scheduled text pipeline stops processing whenever one model call fails on a problematic input. Before implementation, suppliers use incompatible document formats; accounting must prove each extracted total. Which first architectural or operating step MOST directly addresses this case?
- Define mutually understood categories and evaluate classification on a stratified labeled sample.
- Parse the documents into structured text first and evaluate OCR/layout quality before embeddings.
- Profile retrieval latency separately from generation and tune search service design.
- Use an appropriate nonthrowing invocation pattern and handle null or error outcomes per record.
Explanation
Correct Answer
D:The decision addresses the central constraint: a scheduled text pipeline stops processing whenever one model call fails on a problematic input. The most direct next step is to use an appropriate nonthrowing invocation pattern and handle null or error outcomes per record. Before changing production, review both: identify which input produced the failure and compare error handling before and after adjustment.. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
A:This can be appropriate when compliance analysts need to route ticket descriptions among billing, access, privacy, and product issues. It addresses text classification categories, whereas the present limitation is better explained by treating each llm call as infallible makes one transient failure break an entire batch. The target outcome here is that failed rows are quarantined while valid rows finish and are auditable. These author-intended interpretations require independent technical review before operational use.
B:This can be appropriate when operations receives image-heavy scanned forms that must be searchable alongside digital text. It addresses document parsing to text, whereas the present limitation is better explained by treating each llm call as infallible makes one transient failure break an entire batch. The target outcome here is that failed rows are quarantined while valid rows finish and are auditable. These author-intended interpretations require independent technical review before operational use.
C:This can be appropriate when a support bot times out on every query although model completion time alone is acceptable. It addresses cortex search latency, whereas the present limitation is better explained by treating each llm call as infallible makes one transient failure break an entire batch. The target outcome here is that failed rows are quarantined while valid rows finish and are auditable. These author-intended interpretations require independent technical review before operational use.
Question 176
At a university research office, a Snowflake engineering team is undertaking a restricted data discovery project. A scheduled text pipeline stops processing whenever one model call fails on a problematic input. The proposed fix is still hypothetical; staff rotate between roles during the academic year, and research access differs by grant. Which TWO checks provide the most decisive evidence before approving a change?
- Test whether injected text changes the assistant's role
- Identify which input produced the failure
- Compare raw generated payload to JSON schema
- Check chunk length distribution
- Compare error handling before and after adjustment.
Explanation
Correct Answers
B:This observation directly tests the case mechanism: Treating each LLM call as infallible makes one transient failure break an entire batch. In the described setting, it helps determine whether use an appropriate nonthrowing invocation pattern and handle null or error outcomes per record. is justified. These author-intended interpretations require independent technical review before operational use.
E:This complementary observation tests a different part of the same hypothesis and should be read together with the first check before authorizing change. Expected success would be: failed rows are quarantined while valid rows finish and are auditable. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
A:Checking this might help diagnose complete prompt injection, where a helpdesk assistant uses a customer email body as context, and the email orders it to reveal confidential data. It is not one of the two most direct checks for the present case because the leading mechanism is treating each llm call as infallible makes one transient failure break an entire batch. These author-intended interpretations require independent technical review before operational use.
C:Checking this might help diagnose structured outputs schema, where a downstream api rejects model replies because some fields are numeric in one record and strings in another. It is not one of the two most direct checks for the present case because the leading mechanism is treating each llm call as infallible makes one transient failure break an entire batch. These author-intended interpretations require independent technical review before operational use.
D:Checking this might help diagnose recursive text splitting, where a huge support article produces chunks larger than the target embedding and context budgets. It is not one of the two most direct checks for the present case because the leading mechanism is treating each llm call as infallible makes one transient failure break an entire batch. These author-intended interpretations require independent technical review before operational use.
Question 177
At a community-bank technology team, a Snowflake engineering team is undertaking an internal policy assistant pilot. A scheduled text pipeline stops processing whenever one model call fails on a problematic input. Diagnosis is disputed because security rejects any unapproved cross-border data processing; the pilot must serve two departments with different permissions. Which underlying failure mechanism BEST explains the observed limitation?
- Arbitrary slicing can violate size limits and detach concepts from headings.
- A model invocation cannot safely infer mismatched feature names or types.
- Token-based inference cost grows with repeated irrelevant context and overly long responses.
- Treating each LLM call as infallible makes one transient failure break an entire batch.
Explanation
Correct Answer
D:This is the causal hypothesis consistent with the observed limitation: a scheduled text pipeline stops processing whenever one model call fails on a problematic input. Confirm it through both identify which input produced the failure and compare error handling before and after adjustment. before changing the service. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
A:That mechanism could explain a different case about recursive text splitting, but does not most directly account for this specific symptom: a scheduled text pipeline stops processing whenever one model call fails on a problematic input. Here, the evidence to collect is identify which input produced the failure These author-intended interpretations require independent technical review before operational use.
B:That mechanism could explain a different case about model registry signature mismatch, but does not most directly account for this specific symptom: a scheduled text pipeline stops processing whenever one model call fails on a problematic input. Here, the evidence to collect is identify which input produced the failure These author-intended interpretations require independent technical review before operational use.
C:That mechanism could explain a different case about cortex complete budget, but does not most directly account for this specific symptom: a scheduled text pipeline stops processing whenever one model call fails on a problematic input. Here, the evidence to collect is identify which input produced the failure These author-intended interpretations require independent technical review before operational use.
Question 178
At an e-commerce marketplace, a Snowflake engineering team is undertaking a returns reason classification effort. A scheduled text pipeline stops processing whenever one model call fails on a problematic input. The change was trialed on a small dataset; billing and privacy queues need distinct ownership, while some tickets mention several products. Which acceptance condition BEST demonstrates that the intervention worked?
- The retrieval update fixes knowledge omissions without retraining the model.
- High-priority privacy tickets reach the correct reviewer with acceptable recall.
- Failed rows are quarantined while valid rows finish and are auditable.
- Chunks fit budgets without losing essential relationships.
Explanation
Correct Answer
C:This is the observable acceptance condition that would show the intended fix addressed treating each llm call as infallible makes one transient failure break an entire batch. It is stronger than activity counts because it tests the user-facing effect. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
A:This could demonstrate success for model fine-tuning decision, but it would not establish resolution of the current issue: a scheduled text pipeline stops processing whenever one model call fails on a problematic input. Validate failed rows are quarantined while valid rows finish and are auditable. instead. These author-intended interpretations require independent technical review before operational use.
B:This could demonstrate success for text classification categories, but it would not establish resolution of the current issue: a scheduled text pipeline stops processing whenever one model call fails on a problematic input. Validate failed rows are quarantined while valid rows finish and are auditable. instead. These author-intended interpretations require independent technical review before operational use.
D:This could demonstrate success for recursive text splitting, but it would not establish resolution of the current issue: a scheduled text pipeline stops processing whenever one model call fails on a problematic input. Validate failed rows are quarantined while valid rows finish and are auditable. instead. These author-intended interpretations require independent technical review before operational use.
Question 179
At an insurance claims processor, a Snowflake engineering team is undertaking a form intake modernization. A scheduled text pipeline stops processing whenever one model call fails on a problematic input. An auditor will inspect the control design; the new system cannot automatically deny claims, and manual reviewers handle ambiguous scanned content. Which ongoing control MOST directly reduces the original risk?
- Monitor source-index lineage and planned re-embedding.
- Add bounded retries and a dead-letter review path.
- Enforce privacy checks and retention rules at dataset ingress.
- Permit a review/other outcome for ambiguous records.
Explanation
Correct Answer
B:This retains a specific safeguard against recurrence of treating each llm call as infallible makes one transient failure break an entire batch. It complements the immediate intervention, use an appropriate nonthrowing invocation pattern and handle null or error outcomes per record., and allows the team to verify the intended result. These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
A:This is a sensible safeguard for embedding table drift, but it does not target the concrete exposure in this scenario. The current failure is treating each llm call as infallible makes one transient failure break an entire batch. These author-intended interpretations require independent technical review before operational use.
C:This is a sensible safeguard for training example leakage, but it does not target the concrete exposure in this scenario. The current failure is treating each llm call as infallible makes one transient failure break an entire batch. These author-intended interpretations require independent technical review before operational use.
D:This is a sensible safeguard for text classification categories, but it does not target the concrete exposure in this scenario. The current failure is treating each llm call as infallible makes one transient failure break an entire batch. These author-intended interpretations require independent technical review before operational use.
Question 180
At a consumer travel platform, a Snowflake engineering team is undertaking a booking assistant redesign. A scheduled text pipeline stops processing whenever one model call fails on a problematic input. Budget and operational risk are both material; peak demand is highly seasonal; a bad response can trigger a financial adjustment. Which decision MOST appropriately balances the requested outcome with production constraints?
- Audit parse confidence and quarantine illegible pages.
- Permit a review/other outcome for ambiguous records.
- Add bounded retries and a dead-letter review path.
- Use an appropriate nonthrowing invocation pattern and handle null or error outcomes per record.
- Record per-component tracing and latency budgets.
Explanation
Correct Answers
C:This establishes the ongoing safety boundary after the immediate correction. It provides protection while the team checks that failed rows are quarantined while valid rows finish and are auditable. These author-intended interpretations require independent technical review before operational use.
D:This implements the response to the stated failure: Treating each LLM call as infallible makes one transient failure break an entire batch. It must be validated using identify which input produced the failure These author-intended interpretations require independent technical review before operational use.
Incorrect Answers
A:This guardrail primarily addresses document parsing to text and is not the companion intervention the case most urgently requires. The stated issue is a scheduled text pipeline stops processing whenever one model call fails on a problematic input. These author-intended interpretations require independent technical review before operational use.
B:This guardrail primarily addresses text classification categories and is not the companion intervention the case most urgently requires. The stated issue is a scheduled text pipeline stops processing whenever one model call fails on a problematic input. These author-intended interpretations require independent technical review before operational use.
E:This guardrail primarily addresses cortex search latency and is not the companion intervention the case most urgently requires. The stated issue is a scheduled text pipeline stops processing whenever one model call fails on a problematic input. These author-intended interpretations require independent technical review before operational use.