Generative AI Leader Techniques to Improve Generative AI Model Output Practice Test 4 — 15 Questions

Question 1

A developer uses the Gemini API to summarize long articles. The summaries often miss key points from the end of the article. Which technique specifically addresses this length-based loss of information?

Accepted Answer

Break the article into sections and ask the model to summarize each section, then combine. Option D is correct because instructing the model to 'summarize each section and then combine' (chain-of-thought style) helps it process the full document. Option A is wrong because increasing max tokens doesn't change how the model attends to the input. Option B is wrong because truncation worsens the problem. Option C is wrong because a different model with a longer context window could help, but the question is about technique for the current model.

Answer

Increase the max output tokens to 2048

Answer

Truncate the article to the first 2000 tokens

Answer

Use a different model with a larger context window

Question 2

A healthcare startup fine-tunes a model to generate patient education materials. They want to ensure the model never gives medical advice, only information. They add a safety instruction, but the model sometimes still gives advice. What advanced technique should they apply?

Accepted Answer

Apply RLHF with a reward model that penalizes outputs containing medical advice. Option C is correct because reinforcement learning from human feedback (RLHF) with a reward model that penalizes advice can steer the model away from that behavior. Option A is wrong because hard-coded rules may not cover all cases. Option B is wrong because embedding distance is not effective for controlling output content. Option D is wrong because output filtering can block but does not prevent generation of advice in the first place.

Answer

Hard-code a list of prohibited phrases in a post-processing script

Answer

Add a secondary classifier to rewrite any detected advice into general information

Answer

Use semantic similarity to a 'medical advice' embedding and reject if close

Question 3

A company is prompt engineering a model for customer support. They want to reduce hallucination (false information) in responses. Which TWO techniques are most effective? (Choose two.)

Accepted Answer

Implement RAG to retrieve relevant documents for context. Correct options: B and D. B (Retrieval-Augmented Generation) grounds the model in real data. D (specify in system instruction to only use provided facts) instructs the model to rely on context. A (increase temperature) increases creativity, worsening hallucination. C (few-shot examples) helps format but not factuality. E (reduce max tokens) only limits length.

Answer

Provide 3 few-shot examples of conversations

Answer

Reduce max output tokens to 150

Answer

Increase temperature to 1.2

Question 4

A developer is tuning a text-generation model for creative writing. They want the outputs to be more diverse and less repetitive. Which THREE parameters/changes can help? (Choose three.)

Accepted Answer

Increase temperature to 0.9. Increasing temperature to 0.9 raises the randomness of the probability distribution over the vocabulary, making the model more likely to sample less probable tokens. This directly increases output diversity and reduces repetitiveness by flattening the softmax curve, which is a standard technique for creative generation.

Answer

Reduce top-k to 10

Answer

Reduce frequency penalty to 0.0

Question 5

A team is fine-tuning a model for a legal document summarization task. They need to ensure high accuracy and avoid hallucinations. Which TWO approaches should they combine? (Choose two.)

Accepted Answer

Use Retrieval-Augmented Generation to retrieve relevant legal texts. Correct: A and D. A (RAG) provides source material to ground summaries. D (human-in-the-loop validation) catches errors before final output. B (increase temperature) is counterproductive. C (early stopping) addresses overfitting but not factuality. E (character-level tokenization) is not relevant.

Answer

Increase temperature to 1.5 during inference

Answer

Implement early stopping during fine-tuning

Answer

Use character-level tokenization to improve spelling

Question 6

A developer is using Vertex AI PaLM 2 to generate product descriptions. The output is often too verbose and includes irrelevant details. Which technique should the developer apply?

Accepted Answer

Use few-shot prompting with examples of concise descriptions. Option C is correct because the developer needs to constrain the model's output to be concise and relevant. Few-shot prompting provides the model with explicit examples of the desired output format (concise descriptions), guiding it to mimic that style and length. This directly addresses verbosity and irrelevant details without altering the model's fundamental randomness or safety settings.

Answer

Set top_p to 0.1

Answer

Enable safety filters

Answer

Increase temperature to 0.9

Question 7

A company wants to build a customer support chatbot that answers based on internal documentation. They use Vertex AI Search and want to ensure the model only uses retrieved documents. What should they do?

Accepted Answer

Enable grounding with Vertex AI Search. Grounding with Vertex AI Search restricts the model to the retrieved context, preventing it from using internal knowledge. Setting temperature to 0.0 reduces creativity but doesn't enforce retrieval. Fine-tuning is unnecessary, and increasing max tokens doesn't affect retrieval.

Answer

Fine-tune the model on the documentation

Answer

Increase max output tokens

Answer

Set temperature to 0.0

Question 8

A team deployed a fine-tuned model for code generation. After training, the model produces syntactically correct but functionally wrong code. What is the most likely cause?

Accepted Answer

Overfitting to training data. Option D is correct because overfitting to training data causes the model to memorize specific code patterns and syntax from the training set without learning the underlying logic or functional requirements. This results in syntactically correct outputs that fail to generalize to new, unseen coding tasks, producing functionally wrong code despite proper syntax.

Answer

Incorrect prompt format

Answer

Low temperature setting

Answer

Insufficient training epochs

Question 9

A company uses Vertex AI PaLM for code generation. The code often contains security vulnerabilities. Which improvement should be applied?

Accepted Answer

Include a security-focused system instruction. Including a security-focused system instruction (e.g., 'Write secure code that avoids SQL injection') directly guides the model to produce safer output. Increasing temperature worsens security, setting top_k to 1 reduces diversity but doesn't address security, and using Codey alone without instructions may not suffice.

Answer

Set top_k to 1

Answer

Use Codey model instead

Answer

Increase temperature to 0.8

Question 10

After fine-tuning a model on customer support data, the model starts using profanity. What is the most effective mitigation?

Accepted Answer

Enable a safety attribute filter. Enabling a safety attribute filter blocks profanity at inference time. Reducing learning rate may not help, adding profanity as negative examples could help but is less immediate, and increasing temperature increases randomness, potentially worsening the issue.

Answer

Add profanity to training data as negative examples

Answer

Reduce learning rate and retrain

Answer

Increase temperature to reduce confidence

Question 11

A team uses Vertex AI Generative AI Studio to tune a model via RLHF. After tuning, the model outputs are bland. What likely went wrong?

Accepted Answer

Reward model overfits to generic responses. If the reward model overfits to generic responses, it penalizes creativity, leading to bland output. Too many training steps can cause overfitting to the reward model, but the most direct cause is the reward model itself. Low temperature during evaluation is a parameter issue, not tuning. Insufficient data may cause underfitting.

Answer

Insufficient training data

Answer

Too many training steps

Answer

Low temperature during evaluation

Question 12

Which technique allows a model to incorporate real-time data from external APIs?

Accepted Answer

RAG with tool calling. RAG with tool calling enables the model to query external APIs in real-time to retrieve current data. Fine-tuning uses static data, prompt engineering alone doesn't fetch data, and model pruning reduces size.

Answer

Prompt engineering

Answer

Fine-tuning

Answer

Model pruning

Question 13

A model generates biased output. Which technique is least effective?

Accepted Answer

Set frequency penalty to 1.0. Setting frequency penalty to 1.0 reduces token repetition but does not address bias. Fine-tuning on diverse data, adversarial debiasing, and safety filters all directly tackle bias.

Answer

Use adversarial debiasing

Answer

Apply safety filters

Answer

Fine-tune on diverse data

Question 14

To improve factuality in generative AI, which is the best approach?

Accepted Answer

Grounded generation with citations. Grounded generation with citations forces the model to base answers on retrieved evidence, directly improving factuality. Increasing model size may help but not as targeted, reducing length doesn't improve accuracy, and adjusting top_p is unrelated.

Answer

Set top_p to 0.1

Answer

Reduce output length

Answer

Increase model size

Question 15

A healthcare chatbot must avoid hallucinations. Which TWO techniques should the team implement? (Choose two.)

Accepted Answer

Use chain-of-thought prompting. Grounding with a knowledge base ensures responses are based on retrieved facts, and chain-of-thought prompting improves reasoning steps. Higher temperature increases randomness, top_k increases diversity, and frequency penalty affects repetition, not hallucinations.

Answer

Set frequency penalty to 0.0

Answer

Use higher temperature

Answer

Increase top_k to 50