Add KLUE-MRC metrics

[ingyuseong]

Metrics for KLUE-MRC (Korean Language Understanding Evaluation — Machine Reading Comprehension)

Adding metrics for KLUE-MRC.
KLUE-MRC is very similar to SQuAD 2.0 but has a slightly different format which is why I added metrics for KLUE-MRC.

Specifically, in the case of LM Eval Harness, it leverages the scoring script of SQuAD to evaluate SQuAD 2.0 and KorQuAD. But the script isn't suitable for KLUE-MRC because KLUE-MRC is a bit different from SQuAD 2.0. And this is why I added the scoring script for KLUE-MRC.

• All tests passed
• Added a metric card (referred the metric card of SQuAD 2.0)
• Compatibility test with LM Eval Harness passed

References

• KLUE: Korean Language Understanding Evaluation
• KLUE on Hugging Face Datasets
• Add KLUE dataset #2416

[mariosasko]

The metrics API in datasets is deprecated as of version 2.0, and evaulate is our new library for metrics. You can add a new metric to it by following these steps.