Evaluate and compare quality

Overview

Use test datasets to track application quality across runs.

Steps

1. Inside an application, open Monitoring and Evaluations. Create a dataset using the input format shown for that application type.
2. Enter questions and expected outputs. If generating test cases, choose topics, language and count, then review the generated dataset.
3. Run the evaluation, track status and inspect individual cases. Select compatible runs for comparison, and review failures before changing configuration.

Check the result

Evaluation results depend on the dataset and configuration. Training-data export under monitoring prepares data; it does not complete fine-tuning.