Tag
FlexiSLM introduces dynamic frame rate capabilities for speech input and output in spoken language models, outperforming fixed-frame-rate models and enabling controllable inference speed.