Fact · Dictation for work and study
28percent character error rate
On the existing MathSpeech equation benchmark, the authors' models reached a character error rate of 28 percent, against 30 percent for the MathSpeech model, which the authors call comparable.
Checked against its source on October 16, 2026
In the source's own words
“comparable character error rate (CER) results on the MathSpeech benchmark (28% vs. 30%) for the equations conversion”
Speech-to-LaTeX: New Models and Datasets for Converting Spoken Equations and Sentences, arXiv, 2025, Abstract.
Open the original- What it measures
- Character error rate counts the wrong, missing and extra characters in the output against a correct transcript, so a lower figure is better. The MathSpeech benchmark tests isolated equations.
- What it does not show
- Isolated equations, not a whole paragraph of dictated prose. A character error rate of 28 percent is still far from error-free, so every dictated equation needs checking.