ICML 2020 Style & Author Instructions
Important changes from last year are in bold; please read the instructions carefully.
Abstract Submissions: Authors should include a full title for their paper, complete author list, as well as a complete abstract by the abstract submission deadline. Submissions that have “placeholder” (test, xyz, etc.) titles or abstracts (or none at all) at the abstract submission deadline will be deleted. Authors of these types of submissions will not be allowed to submit a full paper. Significant changes to abstracts between abstract submission and paper submission will not be allowed. (A significant change is one which plausibly would affect bidding. Any abstract edits with a character edit distance less than 20 will be accepted; any more significant edits may be reviewed manually and possibly rejected.) Abstracts may be withdrawn up to the paper submission deadline.
Paper Length: Submitted papers can be up to eight pages long (not including references), with unlimited space for references. Any paper exceeding this length will automatically be rejected. All submissions must be electronic, anonymized and must closely follow the formatting guidelines in the templates; otherwise they will automatically be rejected. Accepted papers can be up to nine pages long, not including references, to allow authors to address reviewer comments.
Double-Blind Review: As reviewing is double-blind, papers must not include identifying information of the authors (names, affiliations, etc.). Self-reference or links (e.g., github, youtube) that reveal the authors' identities must be avoided. Papers should not refer to documents that are not available to reviewers; for example, do not redact important citation information to preserve anonymity. Instead, use third person or named reference to this work (e.g., "Xiang et al. showed" rather than "We showed"). Supplementary materials and code should also be anonymized (including, for instance, hardcoded paths or URLs that may give away login identifiers or institution).
Authorship: This year, the author list at the abstract submission deadline will be considered final, and no changes in authorship will be permitted for accepted papers. In general, the author list for submissions should include all, and only, individuals who made substantial contributions to the content of the paper. Each author listed on a submitted paper will be notified of submissions and decisions.
Supplementary Material and Code Submission: Authors have the option of submitting one supplementary manuscript containing further details of their work and a separate file containing code that supports experimental findings; it is entirely up to the reviewers to decide whether they wish to consult this additional material. Supplementary material should be material, created by the authors, that directly supports the submission content. Like submissions, supplementary material must be anonymized. To foster reproducibility, we highly encourage authors to submit code. Reproducibility of results and easy availability of code will be taken into account in the decision-making process. Code may be submitted up to thirteen days after the paper deadline (19 Feb 2020). We encourage authors to make use of this extra week, also to avoid overloading the submission server at the paper deadline. Only reviewers and chairs will have access to submitted code and supplementary material, and will be instructed to delete it after the review process.
Accessibility and Latex Source Submission: Paper and supplements should be written to be accessible to as wide an audience as possible. Because there is better accessibility support--in the form of screen readers--for Latex than PDF, submissions and final papers must include Latex source. Properly anonymized Latex source code for all papers will be required by the code submission deadline (19 Feb 2020). Latex source should be compiled into a single zip file named "latexsource.zip" and uploaded as part of the supplementary material. Additionally, please ensure your paper is readable by all: small fonts should be avoided, figures should not only use colors to distinguish curves, and punctuation should be used in lists. More information on making accessible figures and presentations is available here.
Reproducibility Checklist: To foster reproducibility of machine learning results, authors will be asked to answer all questions from the Reproducibility Checklist during the submission process. The answers, which can be updated before the full paper submission deadline, will be made available to the area chairs and/or reviewers to help them evaluate the submission.
Datasets: If the primary topic of the manuscript is to introduce a new dataset as a community resource, we highly recommend that such datasets must be deposited in a data repository that:
Ensures long term preservation of the data.
Provides a persistent identifier such as Digital Object Identifier or Compact Identifier.
Adheres to Schema.org or DCAT metadata standards.
If--due to the special nature of the data--it cannot be deposited in such repository the justification should be included in the manuscript. For all other datasets we recommend using Zenodo.org or Figshare.com since they facilitate preserving the anonymity of authors during the peer review process. Such datasets should be linked to in the manuscript. In addition the license and/or any access restrictions of the dataset must be described in the manuscript. We strongly encourage using CC0 license whenever possible to maximize the ease of reuse of the shared data.
Dual Submissions: It is not appropriate to submit papers that are identical (or substantially similar) to versions that have been previously published, or accepted for publication, or that have been submitted in parallel to other conferences or journals. Authors submitting more than one paper to ICML 2020 must ensure that these submissions do not overlap significantly with each other in content or results. All such submissions violate our dual submission policy, and the organizers have the right to reject such submissions, and remove them from the proceedings. There are several exceptions to this rule:
Submission is permitted of a short version of a paper that has been submitted to a journal, so long as it has not yet been published at the time of submission to ICML. It is the author’s responsibility to make sure that the journal in question allows dual concurrent submissions to conferences.
Submission is permitted for papers presented or to be presented at conferences or workshops without proceedings (e.g., ICML or NeurIPS workshops), or with only abstracts published.
Submission is permitted for papers that are available as a technical report (or similar, e.g., in arXiv). In this case we suggest the authors not cite the report, so as to preserve anonymity.
Previously published papers with substantial overlap written by the authors must be cited in such a way so as to preserve author anonymity. Differences relative to these earlier papers must be explained in the text of the submission.
Reviewing Criteria: Accepted papers must contain significant novel results that further understanding of machine learning, broadly construed. Results can be theoretical or empirical, quantitative or qualitative, or a combination. Papers should make clear their main claim(s) and offer clear evidence in support of those claims. Results will be judged on the degree to which they have been objectively established and/or their potential for scientific, technological, and/or societal impact. Reproducibility of results and easy availability of code will be taken into account in the decision-making process. The paper review form will be posted here shortly.
Citation and Comparison: Papers are expected to cite all refereed publications relevant to their content, but authors are excused for not knowing about all unpublished work, or work that has been recently posted. Papers (whether refereed or not) appearing less than three months before the submission deadline are considered contemporaneous to ICML submissions; authors are not obligated to make detailed comparisons to such papers (though, especially for the camera ready versions of accepted papers, authors are encouraged to).
Toronto Paper Matching System & OpenReview: ICML uses the Toronto Paper Matching System (TPMS) and OpenReview in order to assign submissions to reviewers and area chairs. Both software packages compute similarity scores between ICML submissions and reviewers’ papers. During the submission process, you will be asked to agree to the use of TPMS and OpenReview for your submission. (Note that OpenReview is used only for matching; your paper will not be made public.)
Code of Conduct: Authors will be asked to confirm that their submissions accord with the ICML code of conduct.
Review Form: You can view the ICML 2020 review form for reference.