How to format a paper that reviewers will love – The IPB style guide.

We want our papers being published and read by others, we want to ship a great idea or a valuable lesson. Reading a paper, however, is an effort for the reader, especially if the reader does not know if the paper is a great one. The first impression is always of key importance and you have to work hard to change the mind of a person afterwards. Thus, ship great work from the beginning on. Style and formatting are important to make a good first impression and it is something that is neglected by quite some authors. You have to think it from the perspective of a potential reviewer: “The authors have not even invested an epsilon effort to create a reasonably looking paper—and now I should to put my valuable time into reviewing it? Fuck it.” Consequently, this style guide provides rules to format a paper in the IPB style. But without further ado, let’s start.

• •

  All our papers follow the same structure, check paper.tex file in this repository. Always stick with it, do not change it.

• •

  If the conference asks for 6 pages, provide 6 pages. There is no reason to have a shorter paper. You can always use the extra space to explain more, to show an additional figure, or to show results. The same holds the other way around. Do not submit papers than exceed the page limit unless it is explicitly allowed (by paying a fee).

• •

  Skip the ”paper organization” part, since it usually does not provide any value (replace sections with the actual titles and you will notice that you write something like ”related work introduces related work”).

• •

  At IROS/ICRA/RSS/CVPR/ICCV the motivation figure is always in the upper right corner. At CVPR/ICCV a figure over the whole page is also common.

• •

  Use macros \etal, …to have a consistent usage.

• •

  Have plots and data to generate these in your paper repository. It is highly recommended to generate plots on the fly through make. (TODO: add example)

• •

  Use abbreviations only when necessary and when you use the term at least two more times. Always define with first occurrence (but generally not within the abstract).

• •

  Do not use citations as nouns or objects, since it makes it harder to read and usually the name also provides context:

  • –

    Bad: [stachniss2004iros] extracts point- wise features using a PointNet, …

  • –

    Good: (The approach of) Wang \etal [stachniss2004iros] extracts point-wise features using a PointNet, …

• •

  Capitalization:

  • –

    Don’t capitalize unless it is the name of a person (e.g., Kalman filter not Kalman Filter)

  • –

    Capitalize headings (”All words four letters or more get capitalized, as do any nouns, pronouns, verbs, or adjectives. Any hyphenated word will require a capital letter after the hyphen in titles (for instance, Pre-University).”). Capitalize the title and all section/subsection names.

  • –

    legend of a plot: always lower-case

• •

  No italics unless you want to interrupt the reading flow (e.g. to emphasize -¿ use not more then two times on a page or paper)

• •

  Non-breaking space (~)

  • –

    In front of formulas:

    the width is computed as~$w=h^2$
    

  • –

    before citations:

    Stachniss \etal~\cite{stachniss2004icra}
    

  • –

    before introducing abbreviations:

    simultaneous localization and mapping~(SLAM)
    

• •

  Use for units a small space \,in front of the unit: 22\,cm, 256\,MB, 1.3\,m^2

• •

  text in math mode should use \text, e.g.,

  $A_{\text{tree}}$
  

• •

  Never \texttt{} except it’s code.

• •

  Make sure the source files for images are in the pics folder as well (unless they are huge).

• •

  Figures should always top: \begin{figrue}[t], use only inline figure if it really makes sense (but Cyrill thinks it never does). A figure in between will always break the reading flow and it is harder to find the text below the figure.

• •

  Always reference figures in the text. Furthermore, you must refer them in the text in the correct order.

• •

  Captions should be descriptive, should explain what is seen. Figure + caption should work without text.

• •

  Ideally, figures can be understood even without reading the caption. Try to keep captions short.

• •

  To get consistent text size in figures: Draw/plot figures in the column width/line width, which you can get with \the\linewidth, which will the print at that place the actual line width in points, \ie, 433.62pt for this document. You can convert this to inch for matplotlib or use it in Inkscape to perfectly draw a figure at the right size. Font size of 8/9 is fine.

• •

  It is advised to generate the plots along with the paper, \ie, you put the data for generating the plot in the repository and generate it with the paper. make makes sure that only the files needed are generated.

• •

  Use booktabs package as it improves the formatting of the table significantly.

• •

  Always use our glorified.bib file and if needed a second bib file called new.bib, in which you put new references.

• •

  Always stick with the style in glorified bib. For example, all key should be [lastname-of-first-author][4-digit-year][journal-or-conference], e.g., stachniss2005rss

• •

  Use the strings to have a consistent style for the proceedings, etc., \ie, booktitle = iros,.

• •

  Use double braces {{...}} around title tags, \ie, {{A Capitalized Title}} to get the correct capitalization in the references.

• •

  Provide a url={} tag with the link to the pdf, since the paper repo will then download the pdf.

• •

  Use plain_abbrv that automatically abbreviates first names, i.e., produces E.X. Ample for Eduarado Xavier Ample.

• •

  Cite the conference/journal version of an arXiv article in the citation if the paper was published at a conference or in a journal.

• •

  Keep the glorified.bib clean!