Skip to content
§5.3

Block Classes

A region could be two things—which class does it get?

✅ NormativeSettled. The requirement holds as written.📄 Page ruleCan be applied from the page alone.Read §5.3 in the RFC ↗

The requirement

Every element MUST have exactly one of the seventeen classes below. Its layout class MUST match the class of the corresponding Markdown element defined in Elements (§3.1). A class MUST NOT be chosen merely to complete a page: where the class is genuinely unclear, the output MUST NOT assert a class the page does not support.

Reading the examples. Each one pairs a source page with the output the rule expects. ✅ Sample: the rule applies. ❌ Counter-sample: it looks like it applies and does not, or a rendering got it wrong. A non-conforming output, in red, is shown only for contrast. … marks content left out of an excerpt.

The seventeen classes

Class Used for
Text One coherent paragraph or prose block
Title The overall document title, on its title or front page
Section-header A separate heading introducing a section or subsection
List-item One marked or hanging list item, including its continuation lines
Caption Text outside normal flow that names or explains a table, picture or formula
Footnote A footnote, source note, method note, or small explanatory note tied to the body
Page-header Repeated or page-specific content in the top margin
Page-footer Bottom-margin content — a page number alone qualifies
Picture A unified photo, diagram, chart, logo, map or molecular structure
Table A ruled or borderless row-and-column structure whose alignment matters
Formula A standalone displayed expression, including its displayed equation number
Checkbox-Selected A selected checkbox mark only — not its label
Checkbox-Unselected An unselected checkbox mark only — not its label
Code Preformatted source, commands, logs, or fixed-width output
Document Index A table of contents, index, or directory-like navigation block
Form Irregular controls or fields arranged for data entry
Key-Value Region A compact label/value group without a regular table grid

How each class is rendered in Markdown is set out in Elements (§3.1).

One class across layout and Markdown

The layout box and the corresponding Markdown represent the same element and MUST use the same class. A producer MUST resolve a disagreement using the printed content and update both representations to agree. A detected label is only a suggestion — see Layout Regions (§5.7).

The class describes the element's role; Markdown syntax alone does not determine it. Text and Footnote both render as paragraphs while keeping their distinct classes. A table or chart Caption uses <caption> inside its object's table HTML and keeps its own Caption class and layout box. A Picture rendered as <table data-source="chart"> keeps the Picture class; a printed Table has the Table class. The renderings prescribed in Elements (§3.1) govern these cases.

The distinctions that decide most cases

Table vs Form A Table has a regular row-column structure. A Form has irregular fields or controls for data entry.
Table vs Key-Value Region A Table repeats rows and columns. A Key-Value Region is a compact label/value group with no regular grid.
Title vs Section-header Title is only the document's overall title on its cover — which sheet that is, is decided once in Document Type (§4.4). A repeated title on a later page is Text or page furniture.
Caption vs Text Proximity is not enough. A Caption names or explains an object and sits outside normal flow. Ordinary nearby prose is Text.
Caption vs Footnote A Caption identifies an object. A Footnote adds a source, method, qualification or note.
Picture vs Formula A complex chemical structure is a Picture. An ordinary standalone mathematical expression is a Formula.
Checkbox The box covers only the square, circle or check mark — never its label.

The diagrams below are synthetic pages shown with their element overlays, one colour per class. There is no diagram yet for Checkbox-Selected, Checkbox-Unselected, Code, Document Index, Form or Key-Value Region; the definitions above govern those.

Title and Text

Title MUST be used only for the document's overall title. Authors, dates, addresses and other front-matter prose MUST stay separate Text elements.

01

✅ SampleThe rule applies

Red = Title · yellow = Textblock-title-text

A cover sheet. Expected boxes: one Title around the title and its Release 4.2.1 line, then three Text elements — the authors, the foundation's address, the date. Each Text element keeps its printed line break.

Source page
Block Classes — Red = Title · yellow = Text. Source page.
Expected output

The Quenra Language Reference Release 4.2.1

Mara Ilvesk and the Quenra core team

Quenra Language Foundation 12 Harrow Lane, Tessby

March 14, 2022

Markdown
# The Quenra Language Reference *Release 4.2.1*

Mara Ilvesk
and the Quenra core team

Quenra Language Foundation
12 Harrow Lane, Tessby

March 14, 2022
Block Classes · Source page

Red = Title · yellow = Text

Block Classes — Red = Title · yellow = Text. Source page.

Section-header and List-item

Each visually distinct Section-header MUST be its own element. Each list item MUST be one List-item element, including its bullet or number and every continuation line.

02

✅ SampleThe rule applies

Dark red = Section-header · blue = List-item · yellow = Textblock-section-list

Example A: a clipped neighbouring column on the left (Text, transcribed only as far as it is visible), a paragraph (Text), and the heading IV. RESULTS (Section-header). Example B: three numbered items, each one List-item whose box takes in its number and its wrapped second line.

Source page
Block Classes — Dark red = Section-header · blue = List-item · yellow = Text. Source page.
Expected output · Example A

se chain ach bead the back- e field for air poten- teractions, nteraction

expectation value of ($$\tau^{n}/n!$$). These moments are used to compare the rates throughout the paper to capture the main features and draw physical information from the distribution.

IV. RESULTS

Markdown
se chain
ach bead
the back-
e field for
air poten-
teractions,
nteraction

expectation value of ($$\tau^{n}/n!$$). These moments are used to compare the rates throughout the paper to capture the main features and draw physical information from the distribution.

IV. RESULTS

Expected output · Example B
  1. die vor dem 1. März 1988 ihren Wohnsitz in der Gemeinde Arvenstadt genommen haben und dort seitdem ununterbrochen gemeldet sind,
  2. deren Zuschuss nach dem 30. April 1999 beginnt und
  3. über deren Antrag auf Förderung vor dem 31. Oktober 2004 entschieden wurde und deren Förderung bis zum 30. Juni 2007 nicht widerrufen worden ist.
Markdown
1. die vor dem 1. März 1988 ihren Wohnsitz in der Gemeinde Arvenstadt genommen haben und dort seitdem ununterbrochen gemeldet sind,
2. deren Zuschuss nach dem 30. April 1999 beginnt und
3. über deren Antrag auf Förderung vor dem 31. Oktober 2004 entschieden wurde und deren Förderung bis zum 30. Juni 2007 nicht widerrufen worden ist.
Block Classes · Source page

Dark red = Section-header · blue = List-item · yellow = Text

Block Classes — Dark red = Section-header · blue = List-item · yellow = Text. Source page.

Footnote and Formula

Each distinct Footnote MUST be its own element. A Formula MUST contain the complete displayed expression and its equation number; inline maths stays part of its Text element.

03

✅ SampleThe rule applies

Dark green = Footnote · grey = Formula · yellow = Textblock-footnote-formula

Example A: two columns of body text and the footnote a) under the left column. Example B: a clipped neighbouring column, a paragraph with inline maths, and footnote 1. Example C: prose interrupted by the numbered displays (1) and (2), each one Formula with its number; the prose above, between and below them is four Text elements. The clipped fragments at the right edge of Example C (Figu, layer, …) have no box in the overlay and are left out.

Source page
Block Classes — Dark green = Footnote · grey = Formula · yellow = Text. Source page.
Expected output · Example A

Maximum Overlap (MaxOv) methods.17 A method similar to the Halberg sampling algorithm performs reweighting in NESS for small systems like birth–death chains.18 A related method based on trajectory reweighting is expected to scale to NESS systems,19 however these

a)Electronic mail: r.voss@caldrivane.example

tions made on a system recorded on the system will fix the pathway. The microscopic change is described by the balance to preserve probability flux. The NESS system is also able to be separated into two parts: dissipative contributions whose sign is fixed by the target

Markdown
Maximum Overlap (MaxOv) methods.<sup>17</sup> A method similar to the Halberg sampling algorithm performs reweighting in NESS for small systems like birth–death chains.<sup>18</sup> A related method based on trajectory reweighting is expected to scale to NESS systems,<sup>19</sup> however these

<sup>a)</sup>Electronic mail: r.voss@caldrivane.example

tions made on a system recorded on the system will fix the pathway. The microscopic change is described by the balance to preserve probability flux. The NESS system is also able to be separated into two parts: dissipative contributions whose sign is fixed by the target
Expected output · Example B

sions fully with casts out of orma-

input $$x$$ pixels to the right and $$y$$ pixels down, once for every $$(x, y)$$ s.t. $$0 \le x, y < f$$. Process each of these $$f^{2}$$ inputs, and interlace the outputs so that the predictions correspond to the pixels at the centers of their receptive fields.

1Assuming efficient batching of single image inputs. The classification scores for a single image by itself take 4.1 ms to produce, which is nearly 30 times slower than the fully convolutional version.

Markdown
sions
fully
with
casts
out of
orma-

input $$x$$ pixels to the right and $$y$$ pixels down, once for every $$(x, y)$$ s.t. $$0 \le x, y < f$$. Process each of these $$f^{2}$$ inputs, and interlace the outputs so that the predictions correspond to the pixels at the centers of their receptive fields.

<sup>1</sup>Assuming efficient batching of single image inputs. The classification scores for a single image by itself take 4.1 ms to produce, which is nearly 30 times slower than the fully convolutional version.
Expected output · Example C

ing layer, these functions compute outputs $$\mathbf{y}_{ij}$$ by

$$\mathbf{y}{ij} = f{ks}({\mathbf{x}{si+\delta i,, sj+\delta j}}{0 \le \delta i, \delta j \le k})$$ (1)

where $$k$$ is called the kernel size, $$s$$ is the stride or subsampling factor, and $$f_{ks}$$ determines the layer type: a matrix multiplication for convolution or average pooling, a spatial max for max pooling, or an elementwise nonlinearity for an activation function, and so on for other types of layers.

This functional form is maintained under composition, with kernel size and stride obeying the transformation rule

$$f_{ks} \circ g_{k's'} = (f \circ g)_{k'+(k-1)s',, ss'}.$$ (2)

While a general deep net computes a general nonlinear function, a net with only layers of this form computes a

Block Classes · Source page

Dark green = Footnote · grey = Formula · yellow = Text

Block Classes — Dark green = Footnote · grey = Formula · yellow = Text. Source page.

More

Rule id layout-block-classes · rule layout blocks