Share of reviews changed
Cumulative since scores came out. Dashed lines show where past years ended up by the final decision.
Raised vs lowered
Cumulative number of reviews whose rating went up or down. In past years about nine in ten changes were raises.
Author responses and reviewer replies
Comments posted on OpenReview during the discussion. We count them; we never store or show their text.
Papers with an author response
Share of scored papers where the authors have posted at least one comment. The dashed line marks where past years ended up (74% to 83% of papers).
Reviewer reply rate
Share of papers with an author response where at least one reviewer has replied to the authors. Shown once at least 100 papers have a visible response.
From which score to which
Rows are the original rating, columns the new one. Above the diagonal is a raise, below it a cut. Newly added reviews are not included.
How far paper averages moved
Change in a paper's average rating, for papers where a reviewer changed a score.
By average score at release
Share of reviews changed in each group; the bar shows raises vs cuts.
By track
Share of reviews changed, highest first. Small tracks move a lot on a few changes.
Compared with past years
Share of reviews whose rating changed between the first public snapshot and the final decision. 2027 is only part way through the discussion.