Data-preserving FigMirror code augmentation dataset.
Each row preserves the original sample ID and includes:
original code and render;
FigMirror-augmented code and render;
side-by-side comparison image;
audit JSON with data/topology preservation flags.
A FigMirror visual-review pass was run after the initial 500-sample augmentation. The repair target was: low-saturation, medium-luminance academic palette; no… See the full description on the dataset page:
https://huggingface.co/datasets/zcahjl3/figmirror-aug500.
