Downloads · 30 days
0
statement-to-tikz
will702/statement-to-tikz
statement-to-tikz is a machine learning model from will702. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
Geometry statement → constraint-solved TikZ (training-free pipeline).
Access
Public
Updated Jul 28, 2026
Repo size
—
Likes
0
Public
Repo makeup
Click a slice to open those files.
.lock118 KB · 62%
At a glance
- License
- mit
- Access
- Public
- Created
- Jul 28, 2026
- Updated
- Jul 28, 2026
- SHA
- c3e1a320
Model card
From the Hugging Face model README
statement-to-tikz
Geometry statement → constraint-solved TikZ (training-free pipeline).
This Hub card mirrors the GitHub project. Source of truth for code is the git repository.
What it does
- Formalize olympiad-style plane geometry text into
GeometryIR(JSON constraints). - Solve point coordinates with SciPy
least_squares(gauge-fixed). - Emit TikZ; verify residuals; optionally compile PDF.
- Fall back to
schematiclayout when exact solve fails.
Quick start
pip install -e ".[dev]"
stt "In triangle ABC, AB = AC, and D is the midpoint of BC. Draw the median AD." -o out/median
Set OPENAI_API_KEY for LLM formalization of arbitrary statements (stt ... --no-offline).
Not a trained model
v1 ships a pipeline + IR + solver, not fine-tuned weights. Fixtures under tests/fixtures/ are the evaluation seeds.
Citation / related work
- MagicGeo (text-guided geometric diagram generation)
- FormalGeo (olympiad geometry formal systems)
- License
- mit
- Created
- Jul 28, 2026
- Updated
- Jul 28, 2026