Across controlled ASR and speech translation experiments, dense feature prepending does not outperform cross-attention in quality and is slightly slower and more memory hungry.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Prepending or Cross-Attention for Speech-to-Text? An Empirical Comparison
Across controlled ASR and speech translation experiments, dense feature prepending does not outperform cross-attention in quality and is slightly slower and more memory hungry.