Releases
The open data Mova is developed and tested on, released together with Mova itself: the texts it reads, the trees it learns from, the problems it solves.
Every version has its own page with its licence, a table of what is inside and the downloads. Each archive unpacks into folders, and every folder has a README that says where the data comes from and how Mova uses it.
Versions
| version | date | what is new | licence | size |
|---|---|---|---|---|
| 0.4 | 2026-10-16 | the school: the books of its books course, 42,759 agreed school trees, the pupils' dictionary of 16,868 words; the trained en model as a download | CC BY-SA 4.0 | 21M + 149M model |
| 0.3 | 2026-10-03 | the tales treebank (2,087 sentences) and contrast sets for bAbI and StepGame | per part | 9.6M |
| 0.2 | 2026-10-01 | the first release: 42 tale books, summaries and questions, treebanks, math scripts | per part | 8.8M |
From the repository, train/get-corpus.sh fetches the current version (or any earlier one: train/get-corpus.sh 0.3) and checks it against its SHA-256.
This page in Markdown → https://md.mova.elamur.ai/free/en/releases/