Fix: correct argument order in MajAtN.compute

[WeiKangda]

Description

In lighteval/src/lighteval/metrics/metrics_sample.py line 1258, the MajAtN.compute method calls:

return self.compute_score(new_model_response, new_doc)

However, compute_score (e.g., MultilingualExtractiveMatchMetric.compute) expects the signature:

def compute(self, doc: Doc, model_response: ModelResponse) -> float

As a result, the arguments are swapped, which cause runtime errors.

To Reproduce

Steps to reproduce the behavior:

1. Install lighteval version 0.12.2.
2. Define a metric using MajAtN with a subclass of MultilingualExtractiveMatchMetric.
3. Call compute on a sample Doc and ModelResponse.
4. Observe a TypeError or incorrect metric values due to the argument order mismatch.

Expected Behavior

MajAtN.compute should pass arguments in the correct order:

return self.compute_score(new_doc, new_model_response)

so that compute_score(doc, model_response) executes correctly.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.