Generative AI Leader Techniques to Improve Generative AI Model Output Practice Test 11 — 15 Questions

Question 1

A retail analytics team uses a Gemini model to answer questions over a large product catalog stored in BigQuery. Answers are sometimes outdated because the model relies on its training data. They want responses to reflect the latest catalog rows and cite the source table. Which technique should they implement?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Use retrieval-augmented generation by querying BigQuery and inserting the retrieved rows into the prompt as grounding context.. Retrieval-augmented generation grounds the model in data fetched at request time, which keeps answers current and allows the source table to be cited. Fine-tuning freezes a stale snapshot, stuffing the whole catalog is impractical, and temperature changes only affect randomness. When answers must track a live system of record, retrieval is the appropriate pattern.

Answer

Increase the model's context window and paste the entire product catalog into every prompt so the model always has all data.

Answer

Fine-tune the model on a recent export of the product catalog so the weights contain current product data.

Answer

Enable a higher temperature so the model can combine its training knowledge with newer patterns and produce fresher answers.

Question 2

A media company uses a Gemini model to produce short news digests from long articles. Editors complain that digests vary in length and sometimes bury the key fact in the middle. The team wants a repeatable, machine-checkable structure for every digest. Which technique should they use?

Accepted Answer

Provide a response schema with defined fields such as headline, key fact, and supporting details, and request JSON output.. A response schema with JSON output turns the digest into a structured object with named fields, so the key fact has a fixed location and length is bounded by the schema. This is directly machine-checkable and repeatable, unlike prose instructions, sampling settings, or style-oriented fine-tuning.

Answer

Fine-tune the model on a corpus of previously published digests written by editors.

Answer

Set the temperature to 0 and the topK to 1 to make output fully deterministic.

Answer

Ask the model to be concise and to put the most important information first in the prompt.

Question 3

A financial services firm is using a large language model to generate quarterly investment summaries from raw market data. The summaries occasionally contain fabricated statistics and sometimes omit key risk factors. The team wants to improve factual accuracy and completeness without retraining the model. Which TWO techniques should they apply? (Choose two.)

Accepted Answer

Retrieval-augmented generation (RAG) using a curated knowledge base of verified financial reports.. Retrieval-augmented generation supplies the model with verified, up-to-date information, reducing fabrication and improving coverage of key points. A fact-verification API adds a validation layer that catches any remaining inaccuracies. Together, they address both hallucination and omission without retraining the model.

Answer

Increasing the temperature setting to encourage more creative wording.

Answer

Fine-tuning the model on a small set of historical summaries to improve style.

Answer

Using a chain-of-thought prompt that asks the model to reason step by step before writing the summary.

Question 4

A team is using a generative AI model to create summaries of customer feedback. The summaries are often too long and include unnecessary details. The team wants to make the summaries more concise without losing key information. Which technique should they use?

Accepted Answer

Use prompt engineering to specify a maximum word count for the summaries.. Prompt engineering with a specified word limit is a straightforward way to control the length of generated summaries. By instructing the model to be concise and setting a maximum word count, the team can achieve shorter outputs that still capture essential information. This approach is flexible and can be adjusted easily without model retraining.

Answer

Increase the temperature parameter to encourage brevity.

Answer

Reduce the model's top-k parameter to limit vocabulary diversity.

Answer

Fine-tune the model on a dataset of short summaries.

Question 5

A support team uses a generative AI assistant to answer customer questions. Agents report that the answers are often too long and include unnecessary background. The team wants the responses to be concise and directly address the question. Which prompt adjustment is most appropriate?

Accepted Answer

Tell the model to answer in two sentences or fewer and to skip background unless the customer asks for it.. The most appropriate adjustment is an explicit prompt constraint that limits answer length and omits background unless requested. Verbosity is a content and instruction-following issue, so a clear directive in the prompt is the right control. Sampling parameters like temperature and top-p do not directly enforce conciseness, and asking for more detail would worsen the problem.

Answer

Add a system instruction that the model should always provide a detailed explanation for every answer.

Answer

Lower the model's temperature to zero so answers become deterministic.

Answer

Increase the top-p value so the model considers more possible words.

Question 6

A software company is using a large language model to generate code snippets from natural language descriptions. The generated code often has syntax errors and does not follow the company's coding standards. Which approach is most effective to improve the quality of the generated code?

Accepted Answer

Fine-tune the model on a dataset of code that adheres to the company's coding standards.. Fine-tuning on a dataset of code that adheres to the company's coding standards is the most effective way to improve the quality of generated code. It directly teaches the model the desired syntax, style, and patterns, reducing syntax errors and ensuring compliance with standards.

Answer

Add a prompt instruction to 'write clean code'.

Answer

Increase the temperature to generate more diverse code solutions.

Answer

Use a larger model with more parameters without fine-tuning.

Question 7

A financial services firm uses a generative AI model to summarize quarterly earnings calls. The summaries must consistently follow a strict format: an executive summary, key financial metrics, and forward-looking statements. The model sometimes omits sections or changes the order. Which technique should they use to enforce the structure?

Accepted Answer

Use controlled generation with a response schema (e.g., JSON schema) to constrain the model's output.. Controlled generation with a response schema constrains the decoding process so that the model can only produce tokens that satisfy a predefined structure. This guarantees that each required section appears and in the correct order, which is essential for consistent financial summaries. Unlike few-shot examples, schema-based control is deterministic and enforceable at the API level, making it the most reliable choice for strict formatting.

Answer

Use few-shot prompting with examples that demonstrate the exact desired format.

Answer

Increase the model's temperature to encourage more structured output.

Answer

Reduce the max output tokens to force the model to include all sections.

Question 8

A marketing team uses a text generation model to draft campaign copy. They want the output to consistently follow a specific brand voice and include a call to action at the end of every draft. Which technique should they apply?

Accepted Answer

Provide a system instruction that defines the brand voice and requires a call to action, plus one or two exemplar drafts.. System instructions plus exemplars are the standard way to enforce tone and mandatory content without training. The instruction defines persistent rules, and examples show the desired pattern. Sampling changes affect randomness, and shorter prompts remove guidance, so neither reliably produces consistent brand voice and a guaranteed call to action.

Answer

Set the temperature to zero so the model always produces the same deterministic wording for every campaign.

Answer

Increase the top-k value so the model samples from a broader set of words and naturally adopts a more distinctive voice.

Answer

Shorten the prompt to only the product name so the model has maximum freedom to express the brand personality.

Question 9

A marketing team is using a text generation model to create ad copy. They notice that the model's output is often bland and lacks creativity. They want to increase the diversity of the generated text while keeping it relevant. Which parameter should they adjust?

Accepted Answer

Temperature. Temperature is the primary parameter that controls randomness in generation. Increasing it allows the model to explore less likely token sequences, resulting in more varied and creative outputs. Other parameters like max output tokens, top-k, and frequency penalty address different aspects such as length, token pool restriction, and repetition, respectively.

Answer

Top-k

Answer

Frequency penalty

Answer

Max output tokens

Question 10

A company is using a generative AI model to create personalized email responses to customer inquiries. The responses sometimes contain factual errors or irrelevant information. The company wants to improve the accuracy and relevance of the responses. Which TWO techniques should they use? (Choose two.)

Accepted Answer

Use prompt engineering to provide clear instructions and examples of desired responses.. Retrieval-augmented generation (RAG) grounds the model's responses in a verified knowledge base, ensuring factual accuracy and relevance. Prompt engineering with clear instructions and examples further guides the model to produce desired outputs. Together, these techniques directly target the issues of factual errors and irrelevance without the overhead of fine-tuning.

Answer

Fine-tune the model on a dataset of customer inquiries and responses.

Answer

Increase the model's temperature to make responses more creative.

Answer

Reduce the model's max output tokens to limit response length.

Question 11

A financial services firm uses a generative AI model to draft client emails. The drafts are accurate but sometimes use an overly casual tone. The firm wants to enforce a consistently formal tone across all drafts. Which approach is most reliable for this requirement?

Accepted Answer

Provide a system instruction that defines the desired formal tone and include a short example of an approved email.. System instructions provide persistent behavioral guidance, and an approved example shows the model the exact tone desired. This combination is more reliable than sampling changes, casual fine-tuning, or brittle post-processing. It directly enforces formality across all drafts and scales without retraining, making it the best fit for the firm's consistency requirement.

Answer

Fine-tune the model on a large corpus of casual emails to broaden its style range.

Answer

Add a post-processing step that replaces casual words with formal synonyms using a fixed dictionary.

Answer

Set the temperature to a high value so the model produces more varied language.

Question 12

A retail company uses a generative AI model to create personalized product recommendations. The model sometimes generates recommendations that include products the company does not sell. Which technique should be used to prevent the model from generating non-existent products?

Accepted Answer

Implement a post-processing filter that checks generated product names against the company's inventory database.. A post-processing filter that cross-references generated product names with the company's inventory database ensures that only real products are recommended. This method is direct, reliable, and does not rely on the model's internal knowledge, effectively preventing hallucinations of non-existent products.

Answer

Increase the model's temperature to allow more creative recommendations.

Answer

Fine-tune the model on a dataset of customer reviews.

Answer

Use a larger context window to include the entire product catalog in the prompt.

Question 13

A team is using a generative AI model to answer customer questions about a complex product. They want to improve the factual accuracy and reduce hallucinations. Which TWO techniques should they apply? (Choose two.)

Accepted Answer

Use a grounded prompt that instructs the model to answer only from provided context and to say 'I don't know' if unsure.. Grounding with RAG supplies authoritative context, and a grounded prompt instructs the model to use only that context and to express uncertainty when appropriate. Together they attack hallucinations at the source: the model no longer relies solely on memorized parameters and is discouraged from fabricating. Other techniques like temperature or top-k affect randomness, not factual grounding, and fine-tuning on generic data does not provide the needed product facts.

Answer

Fine-tune the model on a small set of generic customer service dialogues.

Answer

Reduce the top-k parameter to a very small value.

Answer

Increase the temperature to make the model more confident.

Question 14

A data science team is building a document question-answering assistant on Vertex AI. Users report that answers are sometimes fabricated when the retrieved passages do not contain the answer. Which TWO techniques should the team apply to reduce hallucinations? (Choose two.)

Accepted Answer

Instruct the model to answer only from the provided context and to respond with a fixed phrase such as 'not found in the provided documents' when the context is insufficient.. Grounding instructions with an abstention path and mandatory citations both push the model to rely on retrieved passages and to expose when evidence is missing. Higher temperature and larger output limits increase unsupported text, while fine-tuning stores a stale snapshot without providing query-time provenance. Together, the two prompt and output controls reduce fabrication.

Answer

Raise the maximum output token limit so the model has more space to explain its reasoning and avoid mistakes.

Answer

Increase the temperature so the model explores alternative answers when the retrieved context seems incomplete.

Answer

Fine-tune the model on the full document corpus so all answers are stored in the model weights.

Question 15

A healthcare company is using a generative AI model to draft patient education materials. The model sometimes includes outdated medical advice. The team wants to ensure the content reflects the latest clinical guidelines. They have a database of current guidelines and want to integrate it into the generation process without retraining the model. Which approach should they use?

Accepted Answer

Retrieval-augmented generation (RAG) with the guidelines database as the retrieval source.. Retrieval-augmented generation retrieves relevant, current information from an external database and includes it in the prompt, grounding the model's output in the latest guidelines. This approach avoids retraining and ensures the content is based on verified, up-to-date sources. Other methods like fine-tuning, temperature adjustment, or chain-of-thought do not provide the necessary external knowledge.

Answer

Increasing the model's temperature to encourage more up-to-date responses.

Answer

Fine-tuning the model on the latest guidelines.

Answer

Using a chain-of-thought prompt to ask the model to reason about the latest guidelines.