Skip to main content

​
Get started with Dataset evaluation

Have a dataset ready directly for evaluation? Maxim lets you evaluate your AI’s performance directly without setting up workflows.
1

​
Prepare Your Dataset

Include these columns in your dataset:
  • Input queries or prompts
  • Your AI’s actual outputs
  • Expected outputs (ground truth) Configure Test Run for Datasets
2

​
Configure the Test Run

  • On the Dataset page, click the “Test” button, on the top right corner
  • Your Dataset should already be pre-selected in the drop-down. Attach the Evaluators and Context Sources you want to use. Here’s a polished, documentation-ready version of that line:
  • Map the evaluator’s variables using values from the dropdown, based on your dataset columns. All column types are available for mapping. If a dataset column is of type Variables, you can reference it using dataset.columns["column_name"].
  • Click “Trigger Test Run”
The Dataset appears pre-selected in the drop-down. You can attach the evaluators and context sources (if any) you want to use.Learn more about mapping evaluator variablesConfigure Test Run for Datasets
If you want to test only certain entries from your Dataset, you can create a Dataset split and run the test run on the split the same way as above.