What the Research Paper Reviews
This rubric looks at whether your paper clearly explains a focused research question, the background, methods, results, conclusions, and your own contribution. It also checks that another student could understand how the work was done, that results are compared and supported, and that claims stay within what the data shows.
What We Check
Whether the abstract states the problem, the method, a result with a number, and a conclusion, and whether the paper states one research question whose target the results answer.
Whether the background cites prior work, states what that work leaves unresolved, says what this paper adds, and stays shorter than the paper's own methods and results.
Whether the methods give the data's origin and size, the preprocessing and split with its rule, the model with how it was trained, the settings and how they were chosen, and the code and tools it used.
Whether every result has a named measure, a model and split it came from, a comparison on the same measure, and a figure or table the text actually reads, with both numbers and individual cases.
Whether each conclusion points at a result in the paper, each generalizing claim carries a condition, limitations name a property of the data or method, failures are explained, and overfitting is addressed.
Whether the paper says what the student did and what was provided, what was learned, what would come next, and what is new relative to the cited prior work.
Mistakes We Flag
- Abstract has no finding
- An abstract should summarize the problem, method, main numerical result, and conclusion. Background or detailed procedure alone does not give readers a usable summary.
- Background does not lead to your project
- A long explanation of a field or method can distract from your research question. Use cited prior work to show what remains open and what your paper adds.
- Methods leave out key details
- Readers need the data source and size, preparation steps, data split and assignment approach, training details, and settings so they can understand or repeat the work.
- Results lack a clear comparison
- A number is hard to interpret without a named measure, the data split or denominator, and a baseline or other comparison on the same measure. Tables and figures should also be discussed in the text.
- Claims go beyond the evidence
- A result on one dataset does not by itself show that a model will work everywhere. State the conditions of the data and method, discuss limits and overfitting, and connect conclusions to results in the paper.
How this rubric reads your essay
- Built from published reader guidance — the sources are listed below.
- Scored per section against defined criteria, not general impressions.
- Calibrated for consistent scoring, so a better draft shows up in the score.
References
A summary under 250 words, an introduction with a clearly stated purpose or hypothesis, results that do not overstate conclusions, a discussion of limitations, and methods detailed enough that a different scientist could perform the same experiments.
The journal's scope: hypothesis-driven natural-science work; the example 'We hypothesize machine learning can be used to predict bone spur severity' is listed as unacceptable.
Research Question scored first: a clear and focused purpose, a contribution to the field, and testability; Execution scored on reproducibility of results, appropriate statistical methods, and sufficient data to support conclusions; the interview on understanding the limitations of results and the degree of independence.
The abstract should be 250 words or less and state the problem, the procedure, the key results, and the conclusions, without detailed procedures, tables, or acknowledgements.
The 250-word abstract is written by the student in their own words and must avoid acknowledgements, names of the research institution or mentor, logos or proper names of commercial products, and work done by the mentor; the research plan is a roadmap to review and receive feedback on before beginning.
Procedures must include the methods for data collection and, when applicable, the source of data used, and must delineate what the student will do and what will be done by the mentor.
Every image, graph, table, and chart must be cited; no library research beyond the short introduction; research proposals and incomplete investigations are not eligible; work presented in the student's own words with disclosure of help received.
Judging criteria include the statement of the research problem, logical conclusions relevant to it, whether students recognize their contribution to the field, skill in communicating results, and references stated; the abstract must include the hypothesis and conclusions; presenters should avoid jargon or explain specialized terms.
Tells judges not to reward literature searches presented as original scientific proof or 'what happens if' tests without a supported hypothesis, and to assess what the student actually accomplished versus outside assistance.
Explicit input and output, at least five related-work references, how many training, validation, and test examples and what preprocessing was done, a citation for the dataset's source, primary metrics explained before results, both quantitative and qualitative results, examples of failures, and whether the model overfit.
Claims should match the results in how far they generalize; limitations, data splits, hyperparameters and how they were chosen, and error bars are asked for.
Train/validation/test split details, all preprocessing steps, hyperparameter ranges and selection method, the exact number of runs, and a clear definition of the reported measure.
Meaningful and naive baselines, accuracy on imbalanced classes (a classifier that always outputs the larger class scores 90% while being useless), sequential overfitting when tuning on the test set, and not generalizing beyond the data.
Frequently Asked Questions
What should my abstract include?
Include the problem, method, main result with a number, and conclusion. Keep detailed procedures out of the abstract.
What makes a research question strong for this paper?
It names what you are studying and a relationship, quantity, or property you will measure or compare. Saying only that a model can predict something is not enough without a comparison or a statement that the result could support or weaken.
What details belong in the methods section?
Explain the data origin and size, preprocessing, data split and how records were assigned, the model and training approach, settings and how they were chosen, and the code and tools used. Cite pre-existing code, libraries, models, or datasets you used.
How should I explain my contribution and outside help?
Say which parts you did yourself and what a mentor, provided tool, codebase, data source, or other help contributed. If you say your work is new, connect that statement to cited prior work.
Get your draft scored across all 6 dimension with specific, actionable feedback. Two free reviews per day — no credit card required.