Paper Writing Checklist

Compiled from multiple sources

Note

important problem (inspiring idea) + solid and novel theory + convincing and analytical experiments + good writing = seminal research + excellent paper.

If any of these ingredients is weak, your paper, hence reviewer scores, would suffer.

• Explanatory sections: Related work, Background, Methods. The goal of these sections is to make clear the contributions of the paper and set things up later for the empirical sections.
• Empirical sections: Experiment Setup, Results, Analysis

General

• Number one writing tip: use fewer words. Look for short and simple explanations. Tighten your arguments. Remove extraneous words. Cut filler. Get to the core.

• Start writing the paper early.

• An important concept (eg your proposed solution) should be explained in three ways: as text, as a figure, and as an equation.

• Fill the 8 pages that you are allowed. If you use 7.5 or 7.35, somehow it looks like you weren’t trying or that you don’t really have enough to say.

• Every sentence in your paper falls into 3 categories: 1) High-level: Why are we doing this ? What are its parts ?; 2) Summary: Why might it work ? What does it do ?; and 3) Technical: What are the goals ? What is the process ?. When writing ML papers, you should never be on the second level. You want your writing to be either a high-level description or technical process, no in-between.

Introduction

• Start with the paper with a WHAT figure (known as the “teaser”). This figure shows only two things: 1/ Input and 2/ Output.

Importance (P1)

• What’s the problem?
  • State clearly what problem you are addressing. ::: {.callout-note} This bullet is important to get right for a research paper. The key is to scope the problem correctly. You don’t want a problem statement that’s too broad (“The goal of this paper is to create AGI”) or too narrow (“The goal of this paper is to get XX score on task YY in benchmark ZZ.”). Ref: Dan Fu :::
  • Tell the readers explicitly why they should care about the problem (e.g., applications), ie Why is it interesting and important ?
  • Use your WHAT figure for visual references.
• What have others done?
  • Describe what other solutions (SOTAs) are to your problem.
  • Usually follows a historical trajectory, e.g., classical geometric methods did X, learning-based methods did Y, and recent hybrid methods did Z.

Novelty (P2)

• What’s the gap?: Explain why all these existing solutions are NOT satisfactory (in some aspects). This is important as your work often addresses a specific gap/flaw/drawback of existing work.

Contributions (P3-P5)

• What have you done?:
  • “In this paper, …” at the beginning of this paragraph. It helps provide a clean separation of the past/current work. Structure: Talk about a) Task, b) High-level idea, c) Evaluation
• Make a list of 3 clearly defendable contributions, explicitly stated, so that lazy reviewers can use them in their reviews

Related work

A good previous work section gives the reader insight into the history of the field. It provides a starting point for people to learn about the field but, more importantly, gives them a way to think about it. It frames the history in the context of the current paper. To achieve this, it should organize the work into themes or groups that has commonality. At the same time, it also shows how your contribution differs from others. You want to ultilize other works as a way of explaining your own work.

• Make a spreadsheet of 3 facets of your work and a list of papers

• Structure:

  • P1: Contribution
  • P2: Facet 1
  • P3: Facet 2
  • P4: Contrast (eg we show that a significantly more complex and accurate quantizer can be implemented efficiently at large scale model size.)

• don’t write “Black et al.~ proposed …”. Instead write “Black et al.~ propose…”.

• Use the present tense in this section.

Background

The goal of this section is to set up the foundations so that you can explain your contribution.

• Sufficient technical background with minimal material and notation
• As you write, the methods sections adjust what is included in your background.

Methods

• P1 - P2: High-level intro, then description (with diagram).

• No new notation, if you need it, add to background.

• Sketch methods top down.

• Anything in the method section is expected to be tested.

Experiments

Equations

• Punctuation and equations. Treat equations as you would sentences. That is, use commas and periods as if you were reading the equations like any other text. If a sentence ends with an equation, it should have a period at the end. If an equation is followed by “where x is a variable…”, then the equation should have a comma after it.

• Do not let equations run into the margins or gutter! There are ways to fix this. For an equation like

\[\begin{equation} E(x,y,z) = long equation \end{equation}\]

You have a couple of options. First you can pull part of the equation into the main text We define the objective, \(E(x,y,z)=\)

\begin{equation}
long equation
\end{equation}

The above is often enough if you are only slightly over. If you are more over, then use

\begin{eqnarray}
\lefteqn{E(x,y,z) =} \\
& & part of long equation\\
& &  + more of long equation
\end{eqnarray}

Figures

• Make sure that your caption describes the figure. What should the reader take away? What should they look at? What do the colors mean? Reduce cognitive load on the reader.

• If you have a figure, it should be referred to in the text.

• Text in figures should be big enough to read without zooming.

• Put the figure at the top of the page or column (i.e. \begin{figure}{t} or maybe {tbh}).

• don’t write “In Figure 6 we show..”. Take the “we” out and write “Figure 6 shows….”

Tables

• Tables should show the results from a single experiment

• Caption should be declarative (no I/We), and provide enough information so table is standalone.

Conclusion

• Use present tense

• Provide a concise review of the key insights and results (like an abstract but provides a bit more analysis of the results since the reader has now seen these).

• Phrase your limitations as future work, which shows that, while it is a limitation today, there is a practical path forward to addressing this.

Formatting

• et al. and e.g. should not be italicized.

• You need to use ``quote” — that is use two single quotes in a row to get the proper latex quote.

Grammar

• Pick one tense (present) and stick with it throughout the document. It may sound funny to talk about previous work in the present tense but do it; it usually works.

• Avoid saying things passively. People often write “We can compute the X with Y.” What they mean is “We compute X with Y”. Saying you “can” do it means that you didn’t really do it but maybe could have. If you did it, say you did it. If you didn’t do it, why are you writing this? Same goes for sentences like “The loss can be written as follows:”. Be more direct. “The loss is:”. Get rid of all the extra words and words that indicate that you are unsure or hesitant. Be definitive.

• Be explicit. I often see people write things like “our work is more accurate” or “our method is more robust.” Or they write If you use the word “more,” then you have to tell us more than what? That is, “our work is more accurate than the baseline”. The same applies to the use of “better” — better than what?

• Try to avoid vague words like “provides”, “enables”, “allows”, etc. These don’t tell us anything about the mechanism. How does the method/data provide/enable/allow? This is a general tip to make your science writing more precise.

• People over use “to that end”. When you use it, the “end” has to be really clear. In 80–90% of the cases, the end is not clear.

Citation / Bibliography

• Put your citations in numerical order: [3,9,23,64] not [23,9,64,3].

  NoteWhy is this important?

  It may seem like a small detail to you but it makes it much easier for a reader to refer to the bibliography. It also conveys to the reader that you attend to details. If you attend to a detail like this, you probably attend to more important details as well (like in your code and experiments).

• add

  to your preamble and you never have to sort reference numbers manually again!

• Make sure to cite the most recent/definitive version of a paper; that is, if it appeared at CVPR, don’t cite the arXiv version.

References

• Writing a good scientific paper by Prof. Michael Black

• how i structure introductions to research papers by Prof Dan Fu

• How to write an okay research paper by Prof. Sasha Rush