XCOPA Dataset (merged across languages)

This dataset is a merged version of the XCOPA dataset across candidate languages: it, zh.

It contains both validation and test splits with the following fields:

• premise: The premise of the question.
• choice1: The first choice.
• choice2: The second choice.
• question: The question being asked.
• label: The correct answer (0 for choice1, 1 for choice2).
• idx: The index of the example.
• changed: Whether the example was changed from the original (boolean).
• lang: The language of the example.

Original Dataset

The original XCOPA dataset can be found here.