How Big Is the Impact? After the ICLR Reveal, OpenReview Publishes the Truth

How Big Is the Impact? After the ICLR Reveal, OpenReview Publishes the Truth

ICLR Review Data Leak Sparks Industry-Wide Discussion

Date: 2025-12-01 12:06 Beijing

image

Prominent figures speak out in support of the 20-member team behind the scenes

---

Incident Overview

Recently, the academic world was rocked by the ICLR review data leak incident.

A flaw in the OpenReview platform allowed anyone to retrieve reviewer identities and scores for ICLR 2026 submissions by modifying URL parameters.

For many authors — especially those submitting to top-tier conferences — curiosity about reviewer identity was almost irresistible.

Key Findings from Author Reactions

  • Some discovered their papers, months in the making, were low-rated without clear justification.
  • In some cases, reviewers were close friends or colleagues, revealing potential personal biases.
  • Allegations surfaced of deliberately lowballing scores to eliminate competition or settle personal grudges.
image

---

ICLR’s Immediate Response

On Saturday, ICLR released a notice outlining emergency measures:

  • All Area Chairs reassigned.
  • All review comments and scores reverted to pre-discussion state.

This caused frustration for many:

  • Extensive rebuttals suddenly became obsolete.
  • Late-night discussion records were wiped clean.

The 2026 ICLR process has been described as a turbulent roller coaster.

---

OpenReview Security Analysis

image

OpenReview issued a report confirming automated scraping attacks targeted at ICLR 2026, revealing:

  • 97% of venues unaffected (3,203 total).
  • Of affected venues (~96), half had ≤4 papers queried.
  • In ~50 venues, probing activity was concentrated on a small set of papers.
  • ICLR 2026 suffered large-scale automated attacks, compiling and publishing reviewer identities.

Actions Taken:

  • Patch deployed the same morning vulnerability was detected.
  • Hired external cybersecurity and forensics firms.
  • Conducted code audits and log analysis.
  • Coordinated with platforms, law enforcement, and venues to request data deletion and pursue accountability.

Ongoing analysis means figures and conclusions may change.

---

Confirmed Scraping Attack on ICLR 2026

The attack violated Terms of Service and multiple codes of conduct. OpenReview demands:

  • Immediate deletion of any copied data.
  • No further use or distribution of leaked information.

Broader Context:

  • AI research platforms face increasing targeted attacks.
  • Threats include data scraping, identity forgery, and peer-review sabotage.
  • Calls for integrity, goodwill, and mutual respect remain paramount to maintain the scientific ecosystem.

---

Industry Support for OpenReview

image

Chuang Gan, Principal Researcher at MIT-IBM Watson AI Lab, urged compassion and support:

  • OpenReview is non-profit, run by Andrew McCallum at UMass Amherst.
  • Team of ~20 people supporting most top-tier conferences.
  • Annual funding needs exceed \$2M, raised through effort, not commercialization.

> “We should not blame such a dedicated team for one incident; otherwise, fewer people will take on such responsibilities.” — Chuang Gan

---

Community Discussion Highlights

image

Key viewpoints among commenters:

  • Criticism toward ICLR organizers, not OpenReview.
  • Bugs are inevitable; focus blame on irresponsible reviewers.
  • Suggest delayed transparency, e.g., publishing reviewer names after a year.
image
  • Conferences charging high registration fees should financially support platforms like OpenReview.
image

---

Perspective: Trust and Sustainability in Open Science

Incidents like this show how fragile trust is in open scientific ecosystems.

  • Requires technical resilience.
  • Needs community solidarity.
  • Demands sustainable funding models for critical infrastructure.

Parallel with AiToEarn

Platforms such as AiToEarn官网 are exploring secure, open-source content publishing for creators — including researchers.

  • AI-enhanced content generation.
  • Multi-platform publishing (Douyin, WeChat, YouTube, X, etc.).
  • Built-in analytics and model rankings (AI模型排名).

This approach reflects the same principle: support open contribution while ensuring financial sustainability.

---

References:

image

---

For reprints: Contact the official account for authorization.

Press inquiries & submissions: liyazhou@jiqizhixin.com

Read the original text

Open in WeChat

---

Do you want me to also condense this into a 1-page executive brief for faster reading? That could be useful for press or academic distribution.

Read more

Translate the following blog post title into English, concise and natural. Return plain text only without quotes. 哈佛大学 R 编程课程介绍

Harvard CS50: Introduction to Programming with R Harvard University offers exceptional beginner-friendly computer science courses. We’re excited to announce the release of Harvard CS50’s Introduction to Programming in R, a powerful language widely used for statistical computing, data science, and graphics. This course was developed by Carter Zenke.