Search papers, labs, and topics across Lattice.
Charles University, University of Edinburgh
1
0
2
Text-only and multimodal LLMs achieve only half the precision of humans in labeling speech translation errors, highlighting a significant gap in current evaluation methodologies.
