Databricks GenAI Mock Exam — 8 Medium Questions
Medium-difficulty multiple-choice quiz based on the provided Databricks GenAI mock exam material. Each question targets a different objective from the source.
Try this quiz
Play through the questions and see your score instantly
Ready to test your knowledge?
8 questions · Quick play · Instant results
Make your own quiz videos
Turn any topic into a polished video quiz — with AI-powered questions, voiceover, and animations. No video editing skills needed.
Unlimited quizzes, free to start
Create as many quizzes as you want. Describe your topic and AI builds the questions, answers, and explanations for you.
Customise everything
Pick from stunning templates, tweak colours and fonts, add your branding, and choose between vertical or landscape formats.
Export-ready videos
Download HD videos optimised for TikTok, YouTube Shorts, Instagram Reels, or full-length YouTube — one click, no editing.
No credit card required
Quiz Questions & Answers
Review every prompt, the correct responses, and helpful context to prep for your own run-through.
Question 1: A pipeline expects model output as a JSON object with keys 'summary' and 'priority', but the model returns free-text paragraphs. According to the guidance in the source, what is the BEST first fix?
Raise the model temperature to 1.2 so outputs become more varied
Add more retrieved documents to the prompt to guide formatting
Specify the exact JSON schema, show one or two example outputs, and instruct the model to return only JSON
Replace the chat model with an embedding model to force structure
Question 2: A business needs long incident reports converted into two-sentence briefings. Which NLP task category from the source is most appropriate to evaluate?
Text classification
Summarization
Feature extraction
Token classification
Question 3: A simple RAG chain must fetch related passages and return model text. Based on the chain ordering described in the document, which component order is correct?
Output parser, LLM, retriever, prompt template
LLM, retriever, prompt template, output parser
Prompt template, output parser, retriever, LLM
Retriever, prompt template with context, LLM, output parser
Question 4: A team finds many sampled responses contain invented citations. What does the guide recommend as the BEST next step?
Ignore the problem until formal metrics are defined
Turn the pattern into test cases, guidelines and scorers, and fix the grounding of the prompt
Delete the sampled responses and continue testing
Increase the maximum output tokens for longer responses
Question 5: Chunks from markdown manuals are being cut mid-procedure because fixed 300-character windows are used. What preprocessing change does the source recommend?
Use smaller fixed windows of 100 characters instead
Remove all headings from the documents before chunking
Split on headings and sections, with a maximum size cap and a small overlap
Raise the model temperature to improve chunk quality
Question 6: A RAG app will answer warranty questions. From the source, which documents are MOST appropriate to include as sources?
All company email archives about the product
Social media posts about the product
Old draft versions of policies
The current official warranty policy documents and the approved FAQ
Question 7: Which two practices should be checked on a model card before selecting a hub model for a commercial multilingual application? (Pick the single BEST listed item from the source that mentions both legal and capability checks.)
Whether the model was trained on public web data without further checks
The licence (commercial use allowed), and the supported languages and context length
The colour of the model's logo to ensure branding consistency
Only the number of stars the repo has as a proxy for quality
Question 8: A deployed agent works in notebook but hits permission errors in production. According to the material, what is the MOST likely and correct fix?
Raise the model temperature to avoid permission checks
Re-chunk the documents to reduce index size
Declare needed resources when logging, and grant the endpoint's identity the required privileges
Make the index public so the endpoint can access it