Efficient Streaming LLM for Speech Recognition

Junteng Jia, Gil Keren, Wei Zhou +6

Recent works have shown that prompting large language models with audio encodings can unlock speech recognition capabilities. However, existing techniques do not scale efficiently,…