AI peer review for CS and ML papers
Computer science moves fast, and most rejections cite missing baselines, weak ablations or close prior work the authors did not discuss. NoveltyAxis searches arXiv and the published record for that work and reviews your paper the way a programme committee member would.
- Works on
- PDF or Word: papers, thesis chapters, proposals
- Sources searched
- arXiv, OpenAlex, Semantic Scholar and Crossref
- Price
- Free to start; paid plans from $5 a month; reference checks free
What reviewers in computer science and machine learning check
- Strong, recent and budget-matched baselines.
- Ablations that isolate each component of the method.
- Reproducibility: code, data, hyperparameters, seeds and compute reported.
- Leakage between training and test data, and evaluation on the right splits.
- Closest prior work on arXiv, including preprints from the last few months.
Checklists reviewers use
- Conference reproducibility checklists, such as the NeurIPS paper checklist.
- Artifact evaluation at systems venues.
- Dataset documentation (datasheets) and model cards where relevant.
Before you write: check the idea
The idea check searches arXiv as well as the published record, which matters in a field where much new work appears as preprints first.
What it does not do
- It does not rewrite your work or polish the English.
- It cannot judge whether your field will find the question important; your supervisor can.
- It reads the text of your draft; it does not rerun your analysis or check raw data.
Questions
Does it check my code?
Idea checks can inspect public repositories linked to papers. A paper review reads your manuscript; it does not run your code.
Is my unpublished work private?
Your draft or idea is used only to produce your report and is never used to train AI models.