Written and maintained by CASRAI Editorial Board
Last updated
The International Conference on Learning Representations (ICLR) is a premier academic conference dedicated to advances in representation learning and deep learning. Operating as a core venue for artificial intelligence and machine learning research, ICLR provides a global platform for presenting novel algorithms, theoretical foundations, and empirical evaluations across computer vision, natural language processing, robotics, and reinforcement learning.
For machine learning researchers and institutional administrators navigating machine learning vs artificial intelligence publication venues, ICLR is distinguished by its fully transparent review model hosted on OpenReview. Understanding ICLR submission workflows, reciprocal reviewing obligations, and timeline mechanics is essential for ensuring successful dissemination and maintaining academic standing.
What Is ICLR?
ICLR stands for the International Conference on Learning Representations. Founded to address the rapid evolution of deep neural networks and representation learning, the conference focuses on how computational systems learn meaningful and actionable representations from complex, high-dimensional data. Rather than treating deep architectures as black boxes, research presented at ICLR investigates feature learning, latent space geometry, metric learning, generative modeling, and optimization dynamics.
The conference brings together academic researchers, corporate research scientists, and graduate students. Unlike traditional conferences where reviews remain private between authors and program committees, ICLR pioneered large-scale open peer review in machine learning, publishing anonymous submissions and public reviewer exchanges during the evaluation process.
Who Runs ICLR and What It Covers
ICLR is governed by the International Conference on Learning Representations Board and an appointed annual organizing committee. The conference was established in 2013 by machine learning pioneers Yoshua Bengio and Yann LeCun. The governing board oversees long-term policy, venue selection, and ethical standards, while annual General Chairs and Program Chairs manage the review process, track organization, and technical proceedings.
The conference scope covers theoretical, algorithmic, and applied aspects of learning representations, including:
- Representation learning: Unsupervised, self-supervised, semi-supervised, and supervised representation algorithms.
- Architectures and optimization: Transformer mechanisms, diffusion models, state space models, non-convex optimization, and training dynamics.
- Modality applications: Computer vision, natural language understanding, speech recognition, audio synthesis, biology, and robotics.
- Theory and safety: Generalization bounds, mechanistic interpretability, uncertainty quantification, causal inference, and fairness in learning systems.
In addition to the main technical track of oral and poster presentations, ICLR hosts specialized workshop tracks (such as Tiny Papers and focused domain workshops) and offers a Journal-to-Conference track for papers accepted by partner journals.
How to Submit and Review for ICLR
Submitting research to ICLR requires adhering to a structured multi-stage workflow governed through the OpenReview platform. Researchers must navigate precise formatting, disclosure, and review obligations:
- OpenReview Profile Registration: All co-authors must hold an active OpenReview profile with verified institutional affiliations and conflict of interest declarations prior to the abstract deadline.
- Manuscript Preparation and Formatting: Authors format their manuscript using the official ICLR LaTeX style template. Submissions must adhere strictly to double-blind anonymization guidelines, omitting author names, acknowledgments, and direct citations that compromise anonymity.
- Mandatory LLM Disclosure: Authors are required to formally disclose any significant use of large language models in research ideation, code generation, or paper drafting within the submission portal.
- Abstract and Paper Upload: Authors submit their paper title, abstract, and primary subject areas before the initial abstract deadline (typically mid-September Anywhere on Earth), followed by the full PDF submission prior to the final deadline.
- Reciprocal Reviewer Registration: In accordance with ICLR governance, authors with multiple submissions (or designated laboratory members) must register as reciprocal reviewers, agreeing to review assigned manuscripts if selected by the program committee.
- Bidding and Paper Assignment: Selected reviewers inspect assigned paper pools on OpenReview, submitting bids based on domain expertise and methodological familiarity.
- Interactive Rebuttal and Discussion: Once initial reviews are released in mid-November, authors enter an interactive discussion phase. Authors submit rebuttal responses, upload revised PDF drafts with highlighted modifications, and address reviewer questions directly on OpenReview. Detailed guidance on structuring responses is available in the guide to writing a double-blind rebuttal.
- Area Chair Meta-Review and Final Decisions: Area Chairs and Senior Area Chairs evaluate reviewer scores, author rebuttals, and public comments to formulate meta-reviews, with official acceptance decisions announced in late January.
Strengths and Limitations
Participating in ICLR offers substantial scientific visibility alongside specific operational demands:
- High Transparency: OpenReview provides public visibility into reviewer critiques, rebuttals, and revision histories, creating an educational record for the broader research community.
- Rapid Community Engagement: Because submitted papers and reviews are visible during evaluation, researchers receive immediate community feedback and citations.
- Rigorous Reciprocal Obligations: Authors who fail to fulfill assigned reciprocal reviewing duties or submit substandard reviews risk having their own submitted papers desk-rejected.
- Reviewer Variance: The massive growth in paper submissions creates significant variance in reviewer experience and depth, requiring authors to write robust and defensive rebuttals.
ICLR Compared to Alternative AI Conferences
Researchers typically evaluate ICLR alongside other leading machine learning and artificial intelligence conferences. The following table contrasts key operational characteristics across major venues:
| Conference | Core Scope | Review Platform | Review Openness | Typical Submission Window |
|---|---|---|---|---|
| ICLR | Representation learning, deep architectures, self-supervised learning | OpenReview | Public reviews, rebuttals, and discussion | September (Anywhere on Earth) |
| NeurIPS | Neural information processing, machine learning theory, neuroscience | OpenReview | Double-blind with post-decision public reviews | May (Anywhere on Earth) |
| ICML | Broad machine learning algorithms, optimization, empirical evaluation | OpenReview / CMT | Double-blind with private author rebuttal | January – February |
| AAAI | General artificial intelligence, knowledge representation, reasoning, robotics | OpenReview / CMT | Double-blind two-phase review cycle | August – September |
Frequently Asked Questions
When is the annual ICLR submission deadline?
The ICLR abstract deadline typically occurs in mid-September (Anywhere on Earth), with the full paper PDF deadline following several days later. The author rebuttal and discussion period runs through late November and early December, with final acceptance notifications delivered in late January.
What are the consequences of failing reciprocal reviewer duty at ICLR?
Authors who submit multiple papers to ICLR are invited to serve as reciprocal reviewers. If an author agrees to review but fails to submit timely reviews, submits low-quality placeholder reviews, or ignores warnings from Area Chairs, the program chairs reserve the right to desk-reject the author’s own submitted manuscripts.
Can researchers submit papers to ICLR that are already posted on arXiv?
Yes. Submitting a preprint to arXiv or similar preprint repositories does not violate ICLR dual-submission or anonymity policies. However, authors must avoid explicitly advertising their submissions on social media during the review window to preserve double-blind integrity, and reviewers are instructed to judge the OpenReview submission independently of arXiv drafts.
What is the ICLR policy on generative AI and LLM assistance?
ICLR requires all authors to disclose significant usage of large language models for ideation, writing, or analysis in a dedicated field on the submission form. Reviewers must also disclose any LLM assistance used during review formulation, and fully automated review generation without critical human oversight is strictly prohibited.
Sources
- ICLR Official Governance and Mission — International Conference on Learning Representations (Read November 7, 2026)
- ICLR 2026 Call for Papers and Author Guide — International Conference on Learning Representations (Read November 7, 2026)
- ICLR 2026 Reviewer Guide and Reciprocal Reviewing Policies — International Conference on Learning Representations (Read November 7, 2026)
