Cross-attention in speech-to-text models correlates with saliency-based explanations (Pearson r roughly 0.49-0.75 in the best aggregations) but explains only a minority of the variance, so it should complement, not replace, attribution methods.
Di Gangi, Roldano Cattoni, Luisa Bentivogli, Matteo Negri, and Marco Turchi
1 Pith paper cite this work, alongside 217 external citations. Polarity classification is still indexing.
1
Pith paper citing it
217
external citations · OpenAlex
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Cross-Attention is Half Explanation in Speech-to-Text Models
Cross-attention in speech-to-text models correlates with saliency-based explanations (Pearson r roughly 0.49-0.75 in the best aggregations) but explains only a minority of the variance, so it should complement, not replace, attribution methods.