Enhancing Multilingual Speech Generation and Recognition Abilities in LLMs with Constructed Code-switched Data

Open in new window