AI Paper Review

Get rigorous, actionable feedback on a manuscript alone or ask SAI review to inspect the paper together with its code and data.

How SAI review works

SAI review produces a structured referee report and inline comments covering the research question, methods, evidence, presentation, and reproducibility. Every review becomes a workspace where you can ask follow-up questions and improve the work.

Review modes
  • Document (50 credits). Rigorous review of the document. Documents over 50 pages are priced by length, shown before you submit. We use our best approach based on results from Arena.
  • Document + supplementary materials (100 credits). Also reads the code, data, and other materials, and checks them against the paper.
  • Replicate+ (100 credits minimum). Replicates the paper end to end and verifies its results. Also allows you to extend the work (e.g., applying the same methodology to a new dataset).
What you get
  • Referee report. A structured assessment of the research question, methods, evidence, and presentation, with the main strengths and concerns.
  • Inline comments. Comments tied to specific passages in your document.
  • Reproducibility checks. With code and data attached, SAI review checks that the materials match what the paper reports.
  • Follow-up workspace. Every review becomes a workspace where you can ask questions about the feedback and improve the work.

PDF, TeX, Markdown, and DOCX documents are supported. New accounts get 50 free credits, enough for one document review. See pricing and the FAQ for details.

Paper review examples

Selected from SAI's review and verification of ICML 2026 oral papers.

View all ICML reportsRead ICML replication blog
  1. $\tau^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment

    Victor Barres, Honghua Dong, Soham Ray, Xujie Si, Karthik R Narasimhan

    Referee Report27 review comments
  2. Asymmetric Perturbation in Solving Bilinear Saddle-Point Optimization

    Kenshi Abe, Mitsuki Sakamoto, Kaito Ariu, Atsushi Iwasaki

    Referee Report22 review comments
  3. Benchmarking at the Edge of Comprehension

    Samuele Marro, Jialin Yu, Emanuele La Malfa, Oishi Deb, Jiawei Li, Yibo Yang, Ebey Abraham, Sunando Sengupta, Eric Sommerlade, Michael J. Wooldridge, Philip Torr

    Referee Report27 review comments

What researchers say

“I am amazed at the usefulness of this feedback. We should use it for all our papers.”

Marsha R. Rosner, University of Chicago