plot-figure

Publication-Quality Plot Figures

Render matplotlib or plotly figures as SVG for embedding in Typst reports. Typography matches New Computer Modern so figures integrate seamlessly with the document body.

Matplotlib: Publication Settings

Apply these settings before any fig, ax = plt.subplots() call:

import matplotlib.pyplot as plt
import matplotlib as mpl

mpl.rcParams.update({
    "font.family": "serif",
    "font.serif": ["Computer Modern Roman"],
    "text.usetex": False,          # avoid LaTeX dep; CM font via matplotlib
    "font.size": 9,
    "axes.titlesize": 10,
    "axes.labelsize": 9,
    "xtick.labelsize": 8,
    "ytick.labelsize": 8,
    "legend.fontsize": 8,
    "figure.dpi": 150,
    "axes.spines.top": False,
    "axes.spines.right": False,
    "axes.linewidth": 0.6,
    "xtick.major.width": 0.6,
    "ytick.major.width": 0.6,
    "lines.linewidth": 1.2,
})

Saving as SVG

Always use tight_layout before saving. Never use bbox_inches="tight" alone — it clips axis labels:

fig, ax = plt.subplots(figsize=(4.5, 3.0))  # inches; A4 column ≈ 3.3in

# ... plot here ...

fig.tight_layout(pad=0.4)
fig.savefig("figures/per-comparison.svg", format="svg", dpi=150,
            bbox_inches="tight", pad_inches=0.02)
plt.close(fig)

figsize guide:

• Single-column (A4 margin 25mm): (3.3, 2.5) inches
• Two-column / full-width: (6.5, 3.5) inches
• Side-by-side panel: (6.5, 2.8) — use plt.subplots(1, 2)

Color Palette (semantic, matches arch-diagram)

COLORS = {
    "data":      "#e63946",   # red   — input/output
    "encoder":   "#457b9d",   # blue  — encoding
    "decoder":   "#2a9d8f",   # teal  — decoding
    "attention": "#e9c46a",   # yellow — attention
    "loss":      "#e76f51",   # orange — loss curves
    "baseline":  "#adb5bd",   # gray  — baseline/control
}

Use COLORS["encoder"] for the main model, COLORS["baseline"] for ablated/control conditions. Never use matplotlib's default color cycle in paper figures — it carries no semantic meaning.