This post exists to verify that everything renders correctly. Delete it once you’ve checked.
It has draft: true in the frontmatter, so it appears in quarto preview but won’t show up on the published site or in listings.
Regular paragraph text at the default measure. Here is bold, here is italic, here is both, here is strikethrough, and here is inline_code() which should render in rust.
An external link and an internal link to the CV.
A footnote for an aside that doesn’t belong in the main flow.1 And an inline one.2
A blockquote. It should have a left border, italic text, and the muted grey.
Second paragraph of the same quote.
That was a horizontal rule.
Unordered:
Ordered:
Task list:
Definition list, which is what the CV uses:
Inline math like \(H = -\sum_i p_i \log p_i\) should sit on the baseline without disturbing line height.
Display math:
\[ \mathcal{L}(\theta) = \mathbb{E}_{x \sim \mathcal{D}}\left[ -\log p_\theta(x) \right] \]
An aligned block:
\[ \begin{aligned} \text{Attention}(Q, K, V) &= \text{softmax}\left(\frac{QK^\top}{\sqrt{d_k}}\right)V \\ \text{MultiHead}(Q, K, V) &= \text{Concat}(\text{head}_1, \dots, \text{head}_h)W^O \end{aligned} \]
A numbered, referenceable equation:
\[ \text{KL}(p \parallel q) = \sum_i p_i \log \frac{p_i}{q_i} \tag{1}\]
Referenced as Equation 1.
A matrix:
\[ A = \begin{pmatrix} a_{11} & a_{12} \\ a_{21} & a_{22} \end{pmatrix} \]
A non-executing block, for showing syntax without running it:
def entropy(p, eps=1e-12):
"""Shannon entropy of a probability vector."""
p = p / p.sum()
return -(p * np.log(p + eps)).sum()Another language, to check highlighting is language-aware:
quarto render && quarto publish gh-pagesAn executing chunk with the code visible:
import numpy as np
rng = np.random.default_rng(42)
entropies = rng.beta(2, 5, 1000) * 4
print(f"mean: {entropies.mean():.3f}")
print(f"median: {np.median(entropies):.3f}")
print(f"below 0.5: {(entropies < 0.5).mean():.1%}")mean: 1.169
median: 1.108
below 0.5: 14.2%An executing chunk with the code folded behind a toggle, producing a figure:
import matplotlib.pyplot as plt
plt.rcParams.update({
"figure.facecolor": "none",
"axes.facecolor": "none",
"savefig.transparent": True,
"text.color": "#888",
"axes.labelcolor": "#888",
"xtick.color": "#888",
"ytick.color": "#888",
"axes.edgecolor": "#888",
})
fig, ax = plt.subplots(figsize=(7, 4))
ax.hist(entropies, bins=40, color="#b04a24", edgecolor="none")
ax.set_xlabel("entropy (nats)")
ax.set_ylabel("count")
ax.spines[["top", "right"]].set_visible(False)
plt.show()
Figure 1 should be numbered, captioned, and clickable from that reference.
A chunk with the code hidden entirely, showing only output:
If you can see this line but not the code that produced it, echo: false works.A Markdown table:
| Method | Params | Accuracy | Notes |
|---|---|---|---|
| Baseline | 110M | 84.2% | Standard fine-tuning |
| + pruning | 62M | 83.9% | 44% smaller |
| + distillation | 62M | 85.1% | Best result |
A generated table, to check dataframe rendering:
import pandas as pd
pd.DataFrame({
"layer": range(1, 7),
"mean_entropy": np.round(rng.uniform(0.2, 3.5, 6), 3),
"dead_heads": rng.integers(0, 12, 6),
})| layer | mean_entropy | dead_heads | |
|---|---|---|---|
| 0 | 1 | 3.221 | 4 |
| 1 | 2 | 1.911 | 9 |
| 2 | 3 | 1.261 | 3 |
| 3 | 4 | 2.932 | 3 |
| 4 | 5 | 2.744 | 10 |
| 5 | 6 | 1.851 | 7 |
Referenced as Table 1.
A note callout. Default styling, no title.
Tip callouts take a title attribute.
Warning callout, for caveats that matter.
This one starts folded. Useful for long derivations you don’t want interrupting the main argument.
A claim requiring a source (Vaswani et al. 2017). Another with a page reference (Vaswani et al. 2017, 4). An in-text citation: Vaswani et al. (2017) introduced the architecture.
The bibliography renders below, assuming references.bib exists.
This is margin content. Whether it appears in the margin depends on your layout CSS — currently the margin column is collapsed, so it will likely render inline.
Two side-by-side blocks:
Left column content. Useful for comparing two things.
Right column content. Should sit alongside the left on wide screens and stack on mobile.
This paragraph exists to check the reading measure at your current max-width. A comfortable line length runs somewhere between sixty and seventy-five characters, and the way to judge it is not by counting but by reading: if your eye loses the start of the next line on the return sweep, the column is too wide, and if the text feels choppy and your eye jumps too often, it is too narrow. Read this paragraph at normal speed and see whether anything catches. Then check it again on your phone, where the column will be constrained by the viewport rather than by the max-width, and where the font size matters more than the measure.