Run AI Review
Generate reviewer-style technical reports on a track's papers, paid in credits
AI Review reads each paper the way a careful reviewer would — passage by passage, keeping a running summary of definitions, equations, and claims — and produces a technical report: an overall assessment plus a list of concrete issues (flawed logic, equation inconsistencies, undefined notation, unsupported claims), each anchored to a verbatim quote from the paper.
Papers are sent to an external provider
PaperFox Reviewer2 sends each paper's extracted text to OpenAI — no author information, text only. Refine.ink receives the paper's PDF file. The confirmation dialog states exactly what leaves PaperFox before every run.
Choose a Provider and Model
Two providers are available. PaperFox Reviewer2 runs the same review process on your choice of three OpenAI models — pick per run based on how deep you need the review to be. Refine.ink is a specialist vendor with a single flat-rate review:
| Provider | Model | Cost | Typical 9,000-word paper |
|---|---|---|---|
| PaperFox Reviewer2 | Fast — GPT-5.6 Luna | 15 credits per 1,000 words | ~135 credits |
| PaperFox Reviewer2 | Standard — GPT-5.6 Terra | 35 credits per 1,000 words | ~315 credits |
| PaperFox Reviewer2 | Premium — GPT-5.6 Sol | 70 credits per 1,000 words | ~630 credits |
| Refine.ink | — | 18,000 credits per paper (flat) | 18,000 credits |
Refine.ink is a separate specialist vendor at a premium price point — a proofreading-grade review service. Unlike Reviewer2 (which sends extracted text to OpenAI), it uploads the paper's PDF file to refine.ink for analysis. Its reports appear, share, and bill exactly like the other review services.
Every option is pay-as-you-go: papers are charged one at a time, failed papers are refunded automatically, and a run stops safely if your balance runs out.
Unlike detection, every run produces — and is charged for — a fresh review, even on unchanged papers. Reviews aren't deterministic: running the same model twice gives you two different reports, and both are kept. Use this deliberately — a second opinion from the same model, or a Premium pass over papers Fast flagged.
Reviews take a few minutes per paper
Unlike detection (seconds per paper), a review makes many AI passes over each paper — expect several minutes per paper. The run continues in the background; you'll get an email when it finishes.
Running a Review
- Go to Conferences in the sidebar, then click on your conference
- Navigate to your track and under Submission Management, click "AI Review"
- Pick a Provider from the dropdown — PaperFox Reviewer2 or Refine.ink. With Reviewer2 selected, a Model dropdown picks Fast, Standard, or Premium. The card below shows the selection's rate and paper list — click "Run All", or "Run" on an individual paper's row to review just that paper
- The confirmation dialog shows the estimated cost and the external-data disclosure. Confirm to start
Multi-phase tracks and forms with several file fields get the same Phase and Paper field selectors as detection — see AI Writing Detection.
Reading the Report
Every run of a paper is listed under its row with the provider, model, date, and a verdict badge — No Issues (green), or the issue count (amber, red at 10+). Click a run's magnifier icon to open its full report:
- Overall Assessment — a reviewer-style summary of the paper's strengths and weaknesses
- Issues — each with a title, a tag (technical, logical, or suggestion), the verbatim quote from the paper it refers to, and an explanation of what's wrong
Reports render the vendor's formatting — LaTeX math, tables, and emphasis all display properly, including inside quoted passages.
Use the result filter above the paper list to focus on papers with Issues Found, confirm the No Issues ones, or find papers that were Skipped / Failed.
Reports stay inside PaperFox
Reports are visible to chairs only — unless you share one with the paper's authors (below). The report itself never leaves PaperFox, and no author information is ever sent to a provider.
An AI review is a signal, not a decision
The model can misread notation, miss context, or flag something an expert would accept. Treat the report as a screening aid that tells you where to look — verify every flagged issue against the paper, and never make an accept/reject decision on the report alone.
Every review is kept — a paper's run list spans all providers and models, so you can run Fast across the whole track and Premium on just the borderline papers, then compare all the reports side by side under each paper. The result filter above the list buckets papers by the most recent run of the provider/model selected in the dropdowns.
Sharing a Review with Authors
From the system's point of view a shared AI review behaves like a human review: authors see it on their submission's reviews page under the same release rules as human reviews (your track's review visibility setting — on submit, on release, or on decision).
- Turn on "Share with Authors" at the top of the review page — sharing applies to all of the track's papers at once
- By default each paper shares its latest review — if you re-run a paper, the new report automatically becomes the shared one
- The shared run carries a Shared badge in the paper's run list. To share a different report, click "Share" on that run — re-runs then won't change what authors see
- Turning the switch off hides all reports from authors again
Authors see the report on their submission's reviews page, clearly labeled as AI-generated and presented like a human review — the same overall assessment and issue list you see, but without the verdict badge, provider, model, or run date, which stay chair-only. Reviewers never see AI reviews.
