Paper submission
November 30, 2026
01
Call for Papers · Working draft
Important dates
All dates are provisional. Exact deadline times are to be confirmed.
Reviews released
January 11, 2027
Author rebuttal
January 11–18, 2027
Final decisions
February 8, 2027
02
Review process
Open, public, and adaptive
The conference uses a fully public, real-time review process on OpenReview.
- Submissions are anonymous during review. Author identities are revealed after every final decision, whether accepted or rejected.
- Reviewers remain anonymous permanently.
- Submissions, reviews, author responses, AC discussion, and public comments are visible in real time.
- Anyone with an OpenReview account may comment publicly. Public comments are clearly distinguished from official reviews and do not count as reviews.
Phase 0 Automated screening
Submissions are checked automatically for formatting, anonymity, required sections, file validity, and other mechanical eligibility requirements. There is no human review in this phase.
Phase 1 Initial review
Every remaining paper receives 2 official reviews. Reviewers provide a score and substantive comments without a large collection of separate rating fields.
Unanimous first-round outcomes are recommendations, not automatic decisions. The Area Chair makes an explicit determination before the paper is routed onward.
- Two negative reviews: the AC must either confirm early rejection or rescue the paper into the mixed-review pool. A rescued paper receives 2 additional official reviews, bringing the total to 4, and then continues to rebuttal and public discussion. The AC posts a short public justification for the choice.
- Mixed reviews: 2 additional reviewers are assigned, bringing the total to 4.
- Two positive reviews: the paper normally proceeds directly to rebuttal with its 2 reviewers. The AC may instead require 2 additional official reviews when further scrutiny is warranted, bringing the total to 4 before the paper continues.
Phase 2 Rebuttal and discussion
Continuing papers enter a public rebuttal period. Authors respond to official reviewers, reviewers may revise their assessments, and the community may participate through clearly labeled public comments.
Final Decision
Once a decision is posted, every submission is de-anonymized regardless of outcome. Author identities become public; reviewer identities remain anonymous.
Why adaptive review? It concentrates attention on disputed papers instead of assigning the same review budget to every submission. At a representative 45% escalation rate, the process averages 2.9 reviews per paper versus 5.5 for ICLR 2026—approximately 48% fewer reviews—while preserving AC rescue and additional scrutiny where reviewers disagree. Source ↗