b

bloated-neurips

@preview

NeurIPS-style paper template to publish at the Conference and Workshop on Neural Information Processing Systems

v0.8.0
MIT
Template

Package Information

Last Updated
Minimum Typst Version
0.14.0
Categories
paper
Disciplines
computer-sciencemathematics

Preview

Template preview

1. Get the package

Download the package using the TPIX CLI:

tpix get @preview/bloated-neurips:0.8.0

2. Initialize from template

Create a new project from this template:

typst init @preview/bloated-neurips:0.8.0

Version History

0.8.00.14.0
4c22bb1fb57b...
0.7.00.13.0
d074e6f1142d...
0.5.10.11.1
c22c55651ee3...
0.5.00.11.1
55ef18791094...
0.2.10.10.0
6bf98643cb4d...

Neural Information Processing Systems (NeurIPS)

A template for main track, auxiliary tracks, and workshops.

Configuration

This template exports the neurips2023, neurips2024, neurips2025,
neurips2026 functions with the following named arguments.

  • title: The paper's title as content.
  • authors: A tuple (authors, affls) where authors is an array of author
    dictionaries (keys: name, affl, email, and optional equal) and
    affls is a dictionary mapping affiliation keys to dictionaries (keys:
    department, institution, location, country). The affl key of an
    author entry can be a single string or an array of strings for multiple
    affiliations.
  • keywords: An array of keyword strings for PDF document metadata.
  • date: The publication date, or auto to use the current date.
  • abstract: The content of a brief summary of the paper or none. Appears at
    the top under the title.
  • bibliography: The result of a call to the bibliography function or none.
  • bibliography-opts: A dictionary of named options forwarded to
    set bibliography(...). Defaults to (title: "References", style: "natbib.csl").
  • accepted: If this is set to false then anonymized ready for submission
    document is produced; accepted: true produces camera-ready version. If
    the argument is set to none then preprint version is produced (can be
    uploaded to arXiv).
  • track (neurips2026 only): The submission track for camera-ready copies.
    Valid values are "main" (default), "position", "eandd",
    "creative-ai", and "workshop". Only used when accepted is true.
  • workshop-title (introduced in neurips2026): Title of the workshop. Only
    used when track is "workshop".

The template will initialize your package with a sample call to the
neurips2026 function in a show rule. If you want to change an existing
project to use this template, you can add a show rule at the top of your file
as follows.

#import "@preview/bloated-neurips:0.8.0": appendix, neurips2026

#show: neurips2026.with(
  title: [Formatting Instructions For NeurIPS 2026],
  authors: (authors, affls),
  keywords: ("Machine Learning", "NeurIPS"),
  abstract: [
    The abstract paragraph should be indented ½ inch (3 picas) on both the
    left- and right-hand margins. Use 10 point type, with a vertical spacing
    (leading) of 11 points. The word *Abstract* must be centered, bold, and in
    point size 12. Two line spaces precede the abstract. The abstract must be
    limited to one paragraph.
  ],
  bibliography: bibliography("main.bib"),
  accepted: false,
)

#lorem(42)

#show: appendix

= Technical Details

#lorem(42)

The appendix show rule switches heading numbering to "A.1" style and resets
the heading counter. It can be used instead of (or in addition to) passing
content via the appendix: parameter.

With template of version v0.5.1 or newer, one can override some parts.
Specifically, get-notice entry of aux dictionary parameter of show rule
allows to adjust the NeurIPS 2026 template to a custom workshop as follows.

#import "@preview/bloated-neurips:0.8.0": neurips

#let get-notice(accepted) = if accepted == none {
  return [Preprint.]
} else if accepted {
  return [
    Workshop on Scientific Methods for Understanding Deep Learning, NeurIPS
    2026.
  ]
} else {
  return [
    Submitted to Workshop on Scientific Methods for Understanding Deep
    Learning, NeurIPS 2026.
  ]
}

#let science4dl2026(
  title: [], authors: (), keywords: (), date: auto, /* ... */ body,
) = {
  show: neurips.with(
    title: title,
    authors: authors,
    keywords: keywords,
    date: date,
    /* ... */
    aux: (get-notice: get-notice),
  )
  body
}

Issues

  • There is an issue in Typst with spacing between figures and between figure
    with floating placement. The issue is that there is no way to specify gap
    between subsequent figures. In order to have behaviour similar to original
    LaTeX template, one should consider direct spacing adjustment with v(-1em)
    as follows.

    #figure(
      rect(width: 4.25cm, height: 4.25cm, stroke: 0.4pt),
      caption: [Sample figure caption.#v(-1em)],
      placement: top,
    )
    #figure(
      rect(width: 4.25cm, height: 4.25cm, stroke: 0.4pt),
      caption: [Sample figure caption.],
      placement: top,
    )
    
  • Another issue is related to Typst's inability to produce colored annotation.
    In order to mitigate the issue, we add a script which modifies annotations and
    make them colored.

    ../colorize-annotations.py \
        example-paper.typst.pdf example-paper-colored.typst.pdf
    

    See README.md for details.

  • NeurIPS 2023/2024/2025/2026 instructions discuss bibliography in vague terms.
    Namely, there are no specific requirements. Thus we stick to natbib
    bibliography style since we found it in several accepted papers and it is
    similar to that in the example paper.

  • It is unclear how to render notice in the bottom of the title page in case of
    final (accepted: true) or preprint (accepted: none) submission.