Fact · Voice and AI
334,000pairs
The HyPoradise benchmark for language-model correction of speech recognition contains more than 334,000 pairs of N-best hypotheses and accurate transcriptions.
Checked against its source on October 16, 2026
In the source's own words
“encompassing more than 334,000 pairs of N-best hypotheses and corresponding accurate transcriptions across prevalent speech domains”
HyPoradise: An Open Baseline for Generative Speech Recognition with Large Language Models, Chen Chen et al., arXiv (NeurIPS 2023 dataset track), 2023, Abstract.
Open the original- What it measures
- The benchmark tests whether an external large language model can correct a recognizer's output from its candidate transcripts.
- What it does not show
- A benchmark dataset, not a measure of what a given dictation app does.