Appendix A Details of Modeling

We retrieve top 10 passages and use them as input to mGEN. Gettysburg College, where he was a member of the Lambda Chi Alpha fraternity. We further subsample 50% of the synthetically generated questions. For our multilingual retriever, we split each article into 100-token chunks (Karpukhin et al., 2020), The original passage text file is 29GB, and the total index size is around 129 GB. Both two datasets are under the MIT licence.