single mention · not yet filed
state-switching sequential text/audio generation
In sequential generation mode, the model changes its generation state by generating special control tokens.1 The model switches to audio generation when a special start token is encountered and switches back to the text state when an end token is generated.2
How this definition was written
Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.
In sequential generation mode, LFM2-Audio can change its own generation state by generating special control tokens
— LFM2 Technical ReportWhen a special <|start_of_audio|> is encountered, the model switches to the audio generation state and starts using the RQ-Transformer and audio vocabulary to generate outputs. This continues until a special <|end_of_audio|> token is generated, after which it switches back to the text state
— LFM2 Technical Report
Also called Sequential generation.