Guide
Writing a research paper
What each required section has to do, and the checks a submission has to pass before a reviewer ever sees it.
A paper is not a longer version of your board. The board answers “what did you do” in forty seconds. The paper answers “could someone else do this again” in as many pages as that takes.
Every section below is required. The system checks the ones it can check before your submission reaches a reviewer, so that a reviewer’s time goes to whether the work is sound rather than to whether the abstract is missing.
The sections, and what each one has to do
Title. Sentence case, and specific enough that someone can tell from the title alone what was measured and in what system. If your title would fit fifty other projects, it is not doing its job.
Abstract. What you asked, what you did, what you found, and what it means, in that order. The most common failure is an abstract that describes the procedure and never states the result. Give the number.
Keywords. Three to six. This is where the method goes: if your project applied machine learning to soil chemistry, the discipline is soil chemistry and the keyword is machine learning.
Introduction and background. Why the question matters, and what a reader needs to know before your methods make sense. Not a textbook chapter.
Prior work. What has already been established, and what specifically is missing that your work addresses. A stack of summaries with no argument about the gap is not a literature review. Cite the underlying paper rather than the news article about it.
Methods. The replication test: could a competent stranger repeat this from what is written, without asking you a question? Name equipment, quantities, conditions, sample sizes, and the analysis you ran. State the outlier rule you set in advance.
Results. What you found, presented so the reader can check you. Every figure carries a caption, labeled axes with units, and an explicit statement of what the error bars represent. Every figure is referenced in the body; a figure nobody mentions is a figure nobody reads.
Discussion and limitations. What the results mean, and what they do not support. Limitations stated plainly are a strength. Experienced judges probe them, and a student who has already thought about them answers well.
Conclusion. Short. The claim your data supports, and nothing wider.
Future work. The next experiment, specifically enough that someone could start it.
References. A structured list, in one consistent style.
Contributions statement. Name what each author did and what any mentor did. This is the written form of what independent work scores, and understating a mentor’s role is an integrity problem rather than a presentation one. Write it honestly and it becomes an asset.
What gets checked automatically
These block submission, because they are cheaper for a machine to catch than for a reviewer to report:
- every required section present and above its minimum length
- the abstract within the word limit set by whatever you are submitting to
- at least the minimum number of references
- every figure carrying a caption and alt text, and referenced in the body
- no placeholder text left in the document
- every listed co-author having accepted authorship
- author order explicitly saved
- the contributions statement naming every listed author
What a person checks
Whether the question is clearly stated, whether the methods could be repeated, whether the conclusions follow from the data, whether limitations are honest, and whether the contributions statement matches what the paper describes. Those are judgments, and they stay with a named human.