What problem does it solve?
It removes the time-consuming, error-prone manual work of turning experiment outputs into clean, paper-ready plots and comparison tables.
Core Features & Use Cases
- Data-driven figure generation: creates common ML visuals such as line plots, bar charts, scatter plots, heatmaps, and box/violin distributions directly from JSON/CSV/screen logs.
- Comparison tables and LaTeX-ready outputs: generates standalone LaTeX tables for prior-vs-method or bounds-style comparisons.
- Consistent publication styling: enforces print-friendly typography, DPI, formatting, and reproducibility by generating per-figure scripts and matching style across outputs.
- Preserves existing manual figures: keeps user-provided figures (e.g., hero figures, architecture diagrams) and prepares LaTeX include snippets accordingly.
Quick Start
Use the paper-figure skill to generate all paper figures from your figure plan file PAPER_PLAN.md and your available experiment data, producing publication-quality PDFs plus a figures/latex_includes.tex file for easy copy-paste into your manuscript.