Викторина, связанная с руководством · Жесткий Уровень

Викторина по управлению данными с помощью ИИ

Тесты используют определенные разрешения, происхождение, производные ресурсы, удаление, владение схемой, исправления и постоянную проверку доступа.

Связанные направляющие путиУправление данными ИИ
Вопрос 1 из 8

A dataset may be used for training and evaluation. What must be checked?

Предварительный просмотр каждого вопроса8 вопросы

Ответы и пояснения появляются по мере прохождения теста.

  1. A dataset may be used for training and evaluation. What must be checked?
    • Whether its file opens in the training pipeline
    • Whether its permissions cover each intended use separately
    • Whether both uses require the same model
    • Whether the dataset has a large number of rows
  2. Why does data governance track embeddings and cached responses?
    • They contain no source-related information
    • They are always public
    • They cannot be deleted
    • Derived assets may retain information or dependencies from source data
  3. A deletion procedure removes the uploaded file but not its index copies. What remains unresolved?
    • The model’s user interface
    • The source file’s format
    • Deletion across relevant derived stores and copies
    • The data catalog title
  4. Who should approve changes to data schema and validation rules?
    • An assigned owner responsible for quality and change control
    • Any user who uploads a file
    • The AI model itself
    • The storage vendor only
  5. What does lineage provide?
    • Proof that data use is permitted
    • A guarantee of data quality
    • A substitute for retention rules
    • A way to trace where data and derived outputs came from
  6. A dataset was collected for retrieval but is now proposed for model training. What is needed?
    • Assume access for one use implies every use
    • Review whether its permissions and purpose cover training
    • Delete the data immediately
    • Change the dataset label only
  7. What should a correction procedure specify?
    • Who handles it, how it propagates, and how completion is verified
    • Only where the data catalog is stored
    • The model’s average token count
    • Whether an error appears in a presentation
  8. Why revisit data access and retention when integrations change?
    • Integrations never affect data flow
    • Retention applies only before deployment
    • New services can change who sees the data and how long it persists
    • A new connector automatically narrows access