Search papers, labs, and topics across Lattice.
1
0
3
4
Forget text-first: SODA models show that scaling native audio foundation models with interleaved semantic, acoustic, and text tokens unlocks powerful audio generation and cross-modal capabilities.
