← back to paper
arxiv: 2604.12928 · 2 revisions
MoshiRAG: Asynchronous Knowledge Retrieval for Full-Duplex Speech Language Models