This report synthesises findings from 13 peer-reviewed papers addressing the following research question: Do discrete audio tokens improve cross-lingual speech recognition accuracy compared to mel-spectrograms when fine-tuning large pre-trained models on low-resource languages. 9 claims were extracted from source literature; 9 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 8.7/10. This report is a machine-generated literature synthesis and does not constitute original research.
Research goal: Do discrete audio tokens improve cross-lingual speech recognition accuracy compared to mel-spectrograms when fine-tuning large pre-trained models on low-resource languages?
Autonomous literature synthesis. Automated review score: 8.7/10. Full text and citation available at Assignee Research. Machine-generated literature synthesis. Content is derived from peer-reviewed papers; see individual sources for authoritative data. Automated review score: 8.7/10. Published by Assignee Research (
assignee.net).