B

benchmarking

MultiNativQA is Multilingual Native question-answering (QA) dataset consisting of 64k QA pairs in seven extremely low to high resource languages, covering 18 different topics from nine different regions. Paper: https://arxiv.org/pdf/2407.09823. Project: https://nativqa.gitlab.io

The project uses natural language processing to automatically assess causal diagrams filled out by students based on semantic distance from the model causal diagram provided.