ICLR
The International Conference on Learning Representations (ICLR) is a leading academic conference in the fields of machine learning and deep learning.
Core Facts
Timeline (last 90 days)
Each ICLR reviewer typically needs to independently review 6 to 8 papers.
Faced with the reality of being unable to distinguish the origin of writing, conferences may place greater weight on hard evidence such as open-source code, reproducible experiments, and raw data to filter out slop papers.
A Reddit user initiated a discussion arguing that ICLR may be the first top-tier conference to encounter 'automated research submissions' at scale.
NeurIPS and ICML generally limit the main body of a paper to no more than 8–9 pages (excluding references), and ICLR follows a similar convention.
The ReAct paper, published at ICLR 2023 by Shunyu Yao et al., proposes that thought does not alter the environment and therefore does not produce an observation.
Conferences such as ICML and ICLR have previously revised their policies to explicitly prohibit commercial ticket transfers and require real-name verification for entry.
The error was first discovered by a Reddit user while reviewing the ICLR 2027 author guidelines and style files.
This citation error has existed since the 2019 ICLR template and has been carried over unchanged through each annual version update to the ICLR 2027 template.
The .bib citation adds a "volume 1" designation to the "Deep Learning" textbook that does not actually exist.
In the example .bib file included with the official ICLR template, the textbook "Deep Learning" is cited as "Goodfellow, Bengio, Courville, Bengio", with Yoshua Bengio's name appearing twice.
40 more timeline events
All Facts (20)
ICLR achieves transparency in open peer review through OpenReview
75%VerifiedThe number of submissions to ICLR 2024 surpassed the ten-thousand mark.
75%UnverifiedICLR submission volumes grew from approximately 1,500 in 2019 to surpassing 10,000 in 2022.
70%UnverifiedICLR uses OpenReview's open peer review, where all submitted papers, reviewer comments, author responses, and final decisions are publicly visible, but reviewer identities remain anonymous.
70%UnverifiedOpenReview has been widely adopted by top machine learning conferences such as ICLR.
60%UnverifiedEach ICLR reviewer typically needs to independently review 6 to 8 papers.
50%UnverifiedA Reddit user initiated a discussion arguing that ICLR may be the first top-tier conference to encounter 'automated research submissions' at scale.
50%UnverifiedNeurIPS and ICML generally limit the main body of a paper to no more than 8–9 pages (excluding references), and ICLR follows a similar convention.
50%UnverifiedThe ReAct paper, published at ICLR 2023 by Shunyu Yao et al., proposes that thought does not alter the environment and therefore does not produce an observation.
50%UnverifiedConferences such as ICML and ICLR have previously revised their policies to explicitly prohibit commercial ticket transfers and require real-name verification for entry.
50%UnverifiedThe camera-ready section and FAQ of the ICLR guidelines state that the final version should be consistent with the submitted version (10 pages), contradicting the 9-page limit specified in the formatting instructions.
50%UnverifiedThe ICLR template itself is set to 9 pages.
50%UnverifiedThe .bib citation adds a "volume 1" designation to the "Deep Learning" textbook that does not actually exist.
50%UnverifiedThis citation error has existed since the 2019 ICLR template and has been carried over unchanged through each annual version update to the ICLR 2027 template.
50%UnverifiedIn the example .bib file included with the official ICLR template, the textbook "Deep Learning" is cited as "Goodfellow, Bengio, Courville, Bengio", with Yoshua Bengio's name appearing twice.
50%UnverifiedThe formatting instructions section of the ICLR guidelines states that submissions may have a maximum of 9 pages of body text.
50%UnverifiedWhen page limit rules are ambiguous, choosing the stricter 9-page limit is a safer approach for reducing the risk of rejection.
50%UnverifiedICLR allows community members to participate in review discussions.
50%UnverifiedThe new four-tier system eliminates the neutral option, so reviewers can no longer give a borderline middle score.
50%UnverifiedICLR's new scoring options are: 1 strong reject, 2 weak reject, 3 weak accept, 4 strong accept.
50%