Xiaomi has open-sourced its first native end-to-end voice large model

On September 19th, it was reported that Xiaomi officially open-sourced its first native end-to-end voice model, Xiaomi-Mimo-Audio. Based on an innovative pre-training architecture and over 100 million hours of training data, it has for the first time achieved fest-shot generalization based on ICL in the voice field and observed obvious "emergent" behavior during pre-trainin

g