1 paper · 1 filter
Yuxuan Lou, Ziming Wu, Yaochen Wang +6
Current speech language models generate responses directly without explicit reasoning, leading to errors that cannot be corrected once audio is produced. We introduce \textbf{``Sil…