Skip to main content

​
Purpose of Numeric Difference Eval

  • It evaluate the accuracy of numerical values in model-generated outputs.
  • Unlike semantic or lexical metrics which can overlook numeric discrepancies, NumericDiff ensures that numeric correctness is measured explicitly.

​
Evaluation Using SDK

Click here to learn how to setup evaluation using SDK.
Input & Configuration:
ParameterTypeDescription
Required InputsresponsestrModel-generated text containing the numeric prediction.
expected_textstrGround-truth text with the expected numeric value.
Optional Confignormalized_resultboolWhether to return a normalised score in [0, 1] or absolute difference. If False, raw absolute error is returned. Default: True.
Output:
Output FieldTypeDescription
scorefloatScore between 0 and 1 if normalized_result=True , else absolute difference. Higher is better.
result = evaluator.evaluate(
    eval_templates="numeric_similarity",
    inputs={
        "expected_text": "The Eiffel Tower is a famous landmark in Paris, built in 1889 for the World's Fair. It stands 324 meters tall.",
        "response": "The Eiffel Tower, located in Paris, was built in 1889 and is 324 meters high."
    },
    model_name="turing_flash"
)

print(result.eval_results[0].output)
print(result.eval_results[0].reason)