1 paper
Hao Shi, Yuan Gao, Zhaoheng Ni +4
Large language model (LLM)-based autoregressive speech enhancement (SE) produces natural speech using learned clean-speech priors, but may hallucinate content unsupported by the in…