Search papers, labs, and topics across Lattice.
1
0
3
Just 4 hours of speech data closes the modality gap in LLM-based ASR, rivaling full-dataset fine-tuning and unlocking effective domain adaptation.
