Bài kiểm tra có liên kết hướng dẫn · Cứng Cấp độ

Bài kiểm tra quản trị dữ liệu AI

Kiểm tra các quyền dành riêng cho người sử dụng, dòng dõi, nội dung phái sinh, xóa, quyền sở hữu lược đồ, chỉnh sửa và đánh giá quyền truy cập liên tục.

Đường dẫn hướng dẫn liên quanQuản trị dữ liệu AI
Câu hỏi 1 của 8

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

Xem trước mọi câu hỏi8 câu hỏi

Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.

  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
    • Mô hình AI chính nó
    • 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