​
Experiments

Experiments provide a convenient way to run evaluations on a given data set, typically in CSV format, and analyze the results in a user-friendly interface. This allows you to gain insights and make data-driven decisions based on the evaluation outcomes.

​
Running an Experiment

To run an experiment, you need to follow these steps:
1

Prepare your data set in CSV format, ensuring that it contains the necessary fields for evaluation.

Prepare your data set in CSV format, ensuring that it contains the necessary fields for evaluation.The CSV file should contain the following columns:
idquestionanswerexpectedAnswer
1What is the capital of France?ParisParis
2What is 2 + 2?44
3Is the Earth flat?NoNo
4Who invented the telephone?Alexander Graham BellAlexander Graham Bell
Required Fields:
  • id: A unique identifier for each row
  • question: The input prompt/question to evaluate
  • answer: The actual response/answer to evaluate
Optional Fields:
  • expectedAnswer: The expected/ground truth answer (used for comparison if provided)
The CSV file must include the required fields, while the optional field can be omitted if not needed for your evaluation.
2

Navigate to the Experiments page in the Qualifire platform.

Navigate to the Experiments page in the Qualifire platform.
3

Click on the 'Create New Experiment' button.

Click on the “Create New Experiment” button.
4

Upload your CSV file and specify the evaluation criteria and checks you want to apply.

Upload your CSV file and specify the evaluation criteria and checks you want to apply.
5

Start the experiment and wait for the evaluation process to complete.

Start the experiment and wait for the evaluation process to complete.

​
Analyzing Experiment Results

Once the experiment is finished, you can access the results through the user interface. The UI provides a clear overview of the evaluation outcomes, including any issues or failed checks that were identified. You can explore the results in various ways, such as filtering and sorting based on specific criteria, visualizing trends and patterns, and exporting the data for further analysis. The simplicity and intuitiveness of the UI make it easy to interpret the experiment results and take appropriate actions to improve the quality of your data.