Paper and concept embeddings have been used by several systems to support search and retrieval over the COVID-19 literature. The SPECTER embedding method computes paper embeddings using a SciBERT model [6] pretrained on relatedness signals derived from the citation graph [18]. SPECTER paper embeddings have been shown to successfully capture paper similarity [18] and are available for all papers in CORD-19. Also available for papers in CORD-19 are clinical concept embeddings trained using the JET algorithm [60], relation embeddings trained using SeVeN [28] and network co-occurrence embeddings [63] for biomedical entities computed using CORD-19-on-FHIR. Embeddings capture text similarity and can be used to retrieve similar texts, e.g. the embedding of a query text can be used to retrieve relevant documents from the same embedding space.