AIncorrect
Pretraining
Pretraining is the learning stage that acquires general knowledge from large amounts of unlabeled data.
It is not the stage that measures performance before deployment, so this is incorrect.
A company is planning quality checks along the foundation model lifecycle. Before deploying to production, which stage measures the model's performance and quality with test data or benchmarks to confirm whether it is good enough?
1 / 1
Select an answer
CorrectC
Explanation
A question about choosing the stage that measures performance before deployment.
- 1「Before deploying to production」The stage before deployment
- 2「measures the model's performance and quality」Measuring with tests/benchmarks = evaluation
BIncorrect
Feedback
Feedback is the stage that collects behavior in production and user reactions after deployment and uses them for improvement.
It is not the stage that measures performance before deployment, so this is incorrect.
CCorrect
Evaluation
Correct. Evaluation is the stage that measures the model's performance and quality (accuracy, ROUGE, and so on) with test data or benchmarks before production deployment to confirm whether it meets requirements.
DIncorrect
Deployment
Deployment is the stage itself of rolling out a model that has finished evaluation to production.
It is not the stage that measures performance to confirm whether it is good enough, so this is incorrect.
Key Takeaway
Note the place of the correct answer, evaluation.
- Before production deployment, it measures performance and quality (accuracy, ROUGE, and so on) with test data or benchmarks to confirm whether it meets requirements.
- If there are problems, it returns to retraining or tuning.
Pretraining (learning), feedback (collecting reactions after deployment), and deployment (rolling out) are at different stages.