AI-Powered Paper Review
Beta

Reviewer-grade feedback in 10 minutes

Deep mathematical review with suggested fixes on every issue. Catches subtle errors like normalization gaps, Cauchy-Schwarz misapplications, and compensating errors — with reasoning you can follow.

Currently optimized for mathematics papers. Support for other fields coming soon.

5 credits per section — analyze the whole paper or specific sections.

Uses credits from your ScienceStack subscription. Free tier includes 30 credits/month.

Deep analysis with actionable fixes

Every issue includes reasoning, exact citations, and a suggested fix. Written in a collaborative tone — like feedback from a helpful colleague.

Critical

Heat normalization missing

Section 2.3

The heat kernel satisfies...

The normalization constant doesn't match the standard heat kernel formula. The integral should equal 1, but the stated form integrates to (4πt)^{n/2}. This propagates to Theorem 3.1.

suggestedFix:

Add the normalization factor (4πt)^{-n/2} to the kernel definition.

Significant

Jacobian error in change of variables

Section 4.2

det(J) = r^{n-1}...

The Jacobian determinant has an error in the exponent. Should be r^{n-1} sin^{n-2}(θ₁)... not r^{n}. However, the final answer appears correct — this may be a compensating error.

suggestedFix:

Verify the Jacobian calculation; the result may be correct despite the intermediate step.

Significant

Complex-valued inequality not well-defined

Section 3.1

g^{1/4}(a_m) ≥ |⋯|

g^{1/4}(a_m) is a priori complex-valued, so writing an inequality like g^{1/4}(a_m) ≥ ... is not meaningful without taking absolute values or restricting to real arguments.

suggestedFix:

Replace the inequality by one for absolute values, or add assumption that g is real-valued.

Issues appear as clickable annotations on your paper — jump directly to the problem.

How it works

Upload or open any paper, click “Generate Referee Report”, and get comprehensive feedback in minutes.

Deep Mathematical Analysis

Catches subtle issues like Cauchy-Schwarz misapplications, normalization gaps, and covariance problems that surface-level review misses

Suggested Fixes

Every issue includes an actionable suggestedFix — not just what's wrong, but how to address it

Compensating Error Detection

Recognizes when errors cancel out and the final answer is still correct — avoiding false positives

Collaborative Tone

"I initially stumbled at..." — feedback written like a helpful colleague, not a harsh critic

Reasoning Journey

Shows the confusion and resolution process, so you understand exactly why something was flagged

Better Signal-to-Noise

Filters raw issues down to the ones that matter — fewer false alarms, more real problems

Why we're different

Most AI review tools give surface-level feedback. We built something deeper.

FeatureUsOthers
Suggested fix on every issue
Actionable recommendations, not just problem identification
Collaborative tone
"I initially stumbled at..." instead of terse criticism
Compensating error detection
Recognizes when errors cancel out and final answer is correct
Reasoning journey
Shows the confusion → resolution process
Better signal-to-noise
Filters raw issues to surface real problems
Catches subtle math errors
Normalization gaps, covariance issues, summability problems
Partial

What you get

Comprehensive feedback designed to catch issues before reviewers do.

Turnaround time~10 minutes
Suggested fix on every issueActionable, not just diagnostic
Collaborative tone"I initially stumbled...", "What I checked..."
Compensating error detectionKnows when final answer is still correct
Reasoning journeyShows confusion → resolution process
Better filteringFewer false positives, more real issues

Iterate on your drafts as many times as you need before submission.

Simple credit-based pricing

Referee reports use credits from your ScienceStack account.

5 credits

Per section analyzed

  • Analyze whole paper or specific sections
  • Only critical & significant issues reported
  • Inline annotations on your paper
  • Exact text citations
  • Suggested fixes on every issue
View Plans

Frequently Asked Questions

Ready to improve your paper?

Upload your paper to your library, then use /math_review in the chat to generate a referee report.

    AI Referee Report - Automated Paper Review | ScienceStack