nearly 50%
“Speak's signal-production time was cut by nearly 50%”
As published on labelbox.com. Captured by usedby on Oct 7, 2026.
What happened
Speak, a language learning app, uses Labelbox to produce expert-graded signal that trains and evaluates its speech recognition and LLM-based tutoring models. It runs an automatic labeling loop that evaluates its speech systems daily on production data and retrains the models.
Summary written by usedby from the source page, in English. The figures are those of Labelbox and Speak, not ours.
- 35%“we saw model accuracy improvements of 35% after using Labelbox”
- 2x“what effectively translated into a 2x increase in speed in terms of accelerating model development”
Increased efficiency: Speak's signal-production time was cut by nearly 50%, allowing it to release new features and languages at a faster pace.
At the highest level, we now have a way to tackle something that used to be very painful—an automatic labeling loop that lets us evaluate our speech systems daily on production data.
What the story claims, and what we checked
We compared the story with its live page on Oct 7, 2026.
- The figure: nearly 50%CheckedPrinted word for word on the page, near the name of Speak.
- The passage quoted aboveCheckedCopied word for word from the page, near the name of Speak.
- Speak uses LabelboxCheckedConfirmed line. Latest check across sources: Oct 7, 2026.
- The result itselfNot checkedWe quote it; we did not measure it.




