Most approaches to open question answering retrieve evidence passages from a corpus of documents from which they subsequently extract an answer. However, these approaches cannot be easily extended to a multilingual context, where the knowledge source may be in a different language than the original question. To support multilingual QA, some researchers have developed systems that translate questions into English and answers back into the target language, but this approach may suffer from suboptimal recall and error propagation. As such, Asai et al. present CORA, a single QA model that answers questions across many languages even when language-specific knowledge sources are not available and without using translation modules. CORA includes a multi-lingual retrieval module that uses dense question and passage embeddings to retrieve passage across languages and a generation module that provides answers in the target language based on the retrieved multilingual passages.