Search papers, labs, and topics across Lattice.
Affiliation:
1
0
2
High-performing ASR models can produce accurate transcripts even from contradictory audio, revealing a troubling disconnect between benchmark scores and real-world performance.
