Search papers, labs, and topics across Lattice.
1
0
3
Ditch the text: WavSLM shows you can train a competitive speech language model using only distilled WavLM representations, unlocking a simpler, single-stream generative pretraining paradigm for speech.
