AI Tools: MLflow 3 GenAI Prompt Registry

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-tools-ai-mlflow-3-genai-prompts-r10a4-creators

1. In MLflow 3, what does it mean that prompts are treated as 'first-class artifacts'?

   1. Prompts receive the same treatment as models and other ML artifacts, including versioning and lineage tracking
   2. Prompts are converted to binary format for faster retrieval
   3. Prompts are stored in a special encrypted database that only MLflow can access
   4. Prompts are automatically optimized by MLflow's AI before being stored

2. You discover a typo in a prompt currently in production. What is the correct way to fix it in MLflow 3?

   1. Edit the existing prompt version directly and save
   2. Delete the old version and create a fresh one
   3. Use the rollback feature to revert to an earlier version
   4. Create a new version of the prompt, then promote it through the alias system

3. What is the purpose of aliases in MLflow 3's prompt registry?

   1. To automatically generate prompt variations for testing
   2. To encrypt sensitive prompt content
   3. To provide stable reference names (like 'staging' or 'production') that can point to different versions over time
   4. To compress prompt files to save storage space

4. When should you promote a prompt from staging to production in MLflow 3?

   1. Whenever you have a new version ready
   2. When the prompt passes a basic syntax check
   3. After receiving positive user feedback for 24 hours
   4. When the staging version's evaluation scores are higher than the current production version

5. Which task can MLflow 3's AI capabilities perform for your prompts?

   1. Review your prompt for tone, clarity, and potential bias
   2. Automatically generate A/B test variants based on performance data
   3. Rewrite your prompt to be more engaging without human oversight
   4. Register prompt versions and track which evaluation results belong to which version

6. Why does MLflow 3 not fully automate the human prompt review process?

   1. Because the system lacks sufficient storage for review comments
   2. Because evaluating prompt quality requires subjective judgment that AI cannot reliably replicate
   3. Because human review is optional and only recommended
   4. Because review would violate open-source licensing

7. What happens to traceability if you edit a prompt version in-place rather than creating a new version?

   1. Traceability is broken because the version history no longer reflects what existed at each point
   2. In-place edits are actually required for better performance
   3. Nothing changes—edits are automatically logged
   4. The system automatically creates a backup copy

8. Which statement best describes MLflow 3's evaluation tracking?

   1. Evaluations are only visible in the MLflow web interface, not via API
   2. Evaluations are automatically deleted after 30 days to save space
   3. Each evaluation is tied to a specific prompt version, creating a direct connection between test results and the exact prompt tested
   4. Evaluations are stored separately from prompts with no linkage

9. A team member suggests using MLflow to determine which UI layout works better for their AI-powered app. How should you respond?

   1. MLflow can automatically run UI tests and select the winner
   2. MLflow will handle this if you configure the right integration
   3. MLflow's AI can simulate user behavior to test layouts
   4. MLflow cannot resolve UI/UX questions—this requires separate UX testing tools and human evaluation

10. What is the 'prompt registry' in MLflow 3?

    1. A database that stores all prompts used across the internet
    2. A visual interface for designing prompt templates
    3. A centralized system for managing prompt versions, evaluations, and deployment aliases
    4. A tool that automatically registers new prompts in external AI APIs

11. What type of testing cannot be replaced by MLflow 3's prompt management features?

    1. Load testing that simulates real production traffic
    2. Evaluation metric calculation
    3. Version comparison testing
    4. Alias-based routing tests

12. What is the relationship between prompt versions and evaluations in MLflow 3?

    1. Evaluations are randomly assigned to versions for statistical analysis
    2. Each evaluation is permanently linked to a specific prompt version ID
    3. Evaluations can only be created before a version exists
    4. Versions and evaluations have no relationship—they're stored independently

13. A developer wants to test a prompt in staging before releasing it to all users. What workflow does MLflow 3 support?

    1. Direct deployment to a staging URL without version control
    2. Upload the prompt to a separate staging account
    3. Create a staging alias pointing to the test version, evaluate it, then promote to production alias if successful
    4. Staging doesn't exist in MLflow—only production and development environments

14. Why must a new version be created even for minor prompt changes?

    1. New versions aren't required—only major changes need them
    2. To allow MLflow to compress the data more efficiently
    3. To comply with legal licensing requirements
    4. To maintain complete traceability of every prompt state over time

15. Which of these is listed in the lesson as something AI cannot do for prompt management?

    1. Replace human review of prompt quality
    2. Tie evaluations to prompt versions
    3. Generate prompt variations automatically
    4. Register new prompt versions