ICLR 2027 · acceptance odds

ICLR 2027 acceptance odds simulator

Enter the acceptance rate you expect. The page then draws 10,000 random batches of papers and shows how often a given average score is safely in, needs a tie-break, or is out. The discussion period has ended, so the page uses the actual score distribution from the public ratings on OpenReview as of November 21, 2026, 19:40 UTC.

New Paper score tracker: email alerts when a rating on your paper changesOpenReview does not tell authors when a reviewer changes a rating without commenting. Track up to two papers; we check every 10 minutes. Track your paper →

Settings

Assumed to stay the same until decisions

%

For reference:

papers

ICLR 2026 averaged about 12 submissions per area chair

What the four numbers mean

Safely in
Fewer papers in the batch score at least as high as yours than there are slots. You are in without a tie-break.
Needs a tie-break
Papers that score strictly higher leave at least one slot open, but there are more papers tied with you than slots left.
Out
Papers that score strictly higher than yours already fill every slot in the batch.
Overall chance
The chance of getting a slot if ties are broken at random: “safely in” plus the share of tie-breaks you would win.

This is a simplified model, not how ICLR decides. Treat the numbers as a rough guide.

  • A batch is your paper plus other papers drawn at random from every scored submission, across all areas. Slots per batch = papers per batch × acceptance rate. When that is not a whole number, it is rounded up or down at random in proportion (12 × 27.4% = 3.29, so about 71% of batches take 3 papers and 29% take 4).
  • ICLR does not hand out acceptances per batch, and area chairs weigh much more than the average score. Papers are ranked here by average score only.
  • The score distribution is the actual one after the discussion period: every scored paper's average rating as of November 21, 2026, 19:40 UTC. By then 14.2% of reviews had changed their rating since scores were released.
  • Withdrawn and desk-rejected papers (17.3% of scored papers as of November 21, 2026, 19:40 UTC) can still be drawn into a batch. They count towards its size, so the number of slots stays the same, but they never take a slot. “Share of papers” and “Top” are shares of all scored papers and leave them out of the count as well.
  • Each run draws 10,000 batches, so results carry about ±0.5 percentage points of sampling noise and shift slightly when you draw again.