Input distributions in production move away from training data. What is this?
Why A is the answer
Data drift
Data drift means inputs changed.
In detail
- Inputs moving away from training data is data drift.
- It needs no labels to detect.
Use case
An illustrative scenario, not a real incident.
A model trained on city users now serves rural users, so the input distribution has moved.
See it in different ways
flowchart TD A["Model issue"] --> B["Skew: training and serving data differ"] A --> C["Data drift: inputs change over time"] A --> D["Concept drift: the relationship changes"] A --> E["Attribution drift: feature importance shifts"]
mindmap
root((Decay))
Data drift
Inputs changed
Concept drift
Relationship changed
Skew
Training differs from serving
Attribution drift
Feature importance movedflowchart LR S0["Training distribution"] --> S1["Production distribution"] --> S2["Compare"] --> S3["Alert"]
Key point: Data drift means inputs changed.
Why the other options are wrong
- B. Model parallelism
- That is about training.
- C. A private endpoint
- That is about access.
- D. One-shot prompting
- That is a prompt technique.