My assessments

Two complete, parallel forms. Either one measures the same thing; they exist so a form used for practice is not the one used for a scored sitting.

Loading…

How the score is produced

Every answer is spoken and scored automatically. Parts A–D have answer keys and are scored deterministically; Part B is scored by phonetic alignment against the target sentence. Parts E and F are judged on content by a language model when one is configured, and on fluency alone when one is not — the report says which.

The 20–80 scale and its CEFR labels are provisional. They are internally consistent and separate ability levels reliably, but they have not yet been validated against human raters, so they are not equivalent to a published test score.