The results of ICLR 2026 have been announced and have shocked the academic community. The number of posts skyrocketed to nearly 20,000, but the score plummeted. The average score decreased from 5.12 to 4.2. Reviewers have complained about the poor quality of the papers and even suspect that some were generated by AI. Why has this academic feast lost its flavor?
ICLR 2026 examination results have been announced!
The number of papers submitted to ICLR this year reached a record high.
According to statistics on the Paper Copilot website, the total number of submitted papers in 2025 was 11,672, while ICLR 2026 received 19,631 submissions. Quantity has increased, but quality has decreased.
The highest score in ICLR 2025 was 10 points, while in ICLR 2026 it was 8.5.
The average score for ICLR 2025 was 5.12 and for ICLR 2026 it was 4.20.
Link: https://papercopilot.com/statistics/iclr-statistics/iclr-2026-statistics/
Many submitters have achieved a score of 0 on the ICLR for the first time in their lives.
Some domestic submitters have had all three of their submissions rejected. One of them tweeted this about this unique life experience:
ICLR 2026 will be held in Rio de Janeiro, Brazil from April 23rd to 27th next year.
Along with NeurIPS and ICML, ICLR is one of the three most influential conferences in the field of machine learning and artificial intelligence research.
ICLR, which stands for International Conference on Learning Representations, was founded in 2012 by Turing Award winners Yann LeCun and Joshua Bengio.
ICLR 2026 average score is nearly 1 point lower
Overall, the average ICLR score decreased by nearly 1 point compared to last year.
2026: The top 30% of submissions received scores of approximately 5.0 points or higher.
2025: The top 30% of submissions received a score of 6.0 points or higher.
Afshin Kadangi, a physician in biomedical engineering, used the API to retrieve 19,129 review comments and summarized some of the takeaways.
Only 1,792 papers (approximately 9%) had an average score of 6 or higher.
One paper (id 12681) had 9 comments.
The average individual score was 4.22, with a standard deviation of 1.86, based on 74,371 comments.
Reviewer Cheems Wang found that in his field, papers always have at least one extreme score, and this year’s scores were certainly more conservative.
Reviewer Yu Su also noticed that the average score was quite low. Only 18 out of 225 papers had an average initial score above 6, just barely meeting the conference’s acceptance criteria.
PhD students at Tsinghua University and postdoctoral researchers at Princeton University found a clear scoring pattern for ICLR 2026.
It appears that the higher the submission ID, the lower the score.
Someone noticed a similar pattern two years ago.
To my surprise, there was a comment on the review site that called it “mentally insane.”
This intense review scene caught the attention of Reddit users.
The original review comment has now been deleted. The reviewers apologized for their offensive comments, but maintained that the issues raised in the paper were correct.
Another ICLR reviewer on Amazon also complained:
Immediately after the peer review comments were published, one of the papers I was responsible for was withdrawn.
Additionally, he’s found himself in similar situations several times this year in multiple top-level conferences.
