Paper Review Studio Preview

Make the research story reviewer-ready.

Upload a paper and get a multi-stage AI review: the manuscript is mapped, independent specialists read it, every finding is checked against the page, and the result is one ordered list of what to fix first.

Independent specialist lenses

Story, novelty, architecture, experiments, writing and figures, and consistency are each read by a separate reviewer.

Page-grounded evidence

A finding has to point at the page it comes from. Claims that something is missing are challenged before they are kept.

An actionable revision order

Issues are deduplicated and ranked, each with an exact task and a condition that tells you when it is done.

The full procedure

Independent reviewers first, then verification.

The report is written only after every candidate finding has been checked against the PDF. A review usually takes 3 to 8 minutes.

Map the manuscript

Sections, claims, figures, baselines, results and limitations, and the links from each claim to its evidence.

Run the specialists

Each lens reviews on its own, with extra criteria for a thesis or a project report.

Verify every finding

Page anchors, absence claims, causal reasoning, coverage and priority are all challenged.

Synthesize the order

One list of what to revise, most important first, in the voice of a professor who wants the paper accepted.

The backbone of every review: Problem → Gap → Challenge → Mechanism → Evidence → Takeaway.

What it reviews

From a conference paper to a thesis.

Choose the kind of document and, if you have one, the venue you are aiming for. For a revision, add the earlier reviews and the studio audits whether each comment was addressed.

Documents
Publication paper, PhD thesis, Master's thesis, qualifying exam or research proposal, project report
Submission
A new paper, or a revision compared with a previous review
Depth
Full, with every applicable specialist, or quick, with three
Report language
English or Chinese
Input
A manuscript PDF of up to 25 MB

AI-native, on our own stack

Built on Token Factory.

Paper Review Studio runs its reviewers on the open models MarsCompute serves. It is the first application we have built on our own API, and a pattern for the ones that follow.

  • Many models, one review Pick the model that reads your paper, from the Token Factory catalog.
  • Your history is yours Papers and reviews are kept in your signed-in history, and you can delete a review and its files at any time.
  • In preview The studio is in development. Write to us and we will tell you when a place opens.

Review the paper before a reviewer does.

Ask for preview access, or see what else runs on the same stack.