NeurIPS 2025 Workshop on Constrained Optimization for Machine Learning

Reviewer Resources

If you are interested in reviewing for the workshop, please fill out the reviewer nomination form by August 21st.

Guidelines

Thank you for serving as a reviewer for COML 2025!

You will be assigned 3 extended abstracts (4 pages each) to review.

Important Dates

  • Paper assignment: Aug 26, 2025 August 29, 2025
  • Reviews due: September 11, 2025 (AoE)

Preparation:

  • COML 2025 utilizes OpenReview. Please ensure your OpenReview profile is up to date, as most communications will be sent through this platform.
  • Familiarize yourself with the workshop’s scope and topics by reviewing the Call for Contributions.

Code of Conduct

COML follows the reviewer confidentiality policies of NeurIPS 2025 (see NeurIPS 2025 Reviewer Guidelines). Reviewers are expected to maintain the confidentiality of all submissions until they are made publicly available.

COML reviewers are also expected to adhere to the NeurIPS 2025 Code of Conduct and the NeurIPS 2025 LLM Policy for Reviewers. In particular, sharing a submission with any LLM constitutes a breach of confidentiality and is strictly prohibited. While LLMs may be used to assist in writing reviews, reviewers remain fully responsible for the content of their reviews, which must not be generated by an LLM.

The review process is double-blind; therefore, ensure your reviews do not disclose your identity.

Desk-Rejections and Conflicts of Interest

Should you identify any grounds for desk rejection, please note this in the dedicated field in the review form. Grounds include:

  • Constituting previously published work, or otherwise violating our Dual Submission policy.
  • Exceeding the four-page limit or other serious violations of the formatting guidelines.
  • Violating the double-blind policy.
  • Falling significantly outside the workshop’s scope. See the Call for Papers for relevant topics.
    • If it is not immediately clear whether this applies, please flag the concern but still complete the review, in case the submission is deemed relevant by the AC.
  • Not adhering to NeurIPS’ LLM use policies, e.g., if the paper appears to be entirely generated by an LLM.

Note that there is no checklist requirement for COML 2025 submissions.

Conflicts of Interest: While conflicts of interest are automatically managed by OpenReview based on affiliation and collaboration, you must immediately inform your Area Chair if you identify any conflict not captured by the system.

Review Criteria

Reviewers should assess submissions based on the following aspects:

  • Relevance to the workshop.
  • Soundness:
    • Is the idea feasible and well-motivated?
    • Is the submission adequately grounded in prior work?
    • Is the submission scientifically correct?
  • Clarity and Coherence: Is the submission well-written and easy to follow?
  • Impact Potential: Does the submission present an interesting idea that shows promise for future impact? Note: Impact Potential should primarily be considered when recommending submissions for contributed talks, rather than as a key factor for overall acceptance (see calibration note below).

While providing feedback to help improve the submission and its underlying ideas is encouraged, it is not required.

Note for Calibrating Reviews: COML emphasizes early-stage work. Submissions may present preliminary ideas or incomplete results—this alone should not be grounds for rejection. Any relevant and well-supported submission (i.e., scientifically sound), even if incremental, should be accepted. Exceptional submissions that show significant promise for future impact may be recommended for a contributed talk.

Appendices: Reviewers are not obligated to read appendices. The main text should contain all necessary information for a thorough evaluation.

Review Structure

Each review, to be submitted directly in OpenReview, will consist of the following components:

  • A 2–3 sentence summary of the paper — this should be descriptive, not critical.
  • A multiple-choice question assessing the submission’s adherence to the workshop policies (e.g., respecting the 4-page limit — see Desk Rejections).
    • You may submit a placeholder review if you identify any serious violations, since the paper will be desk rejected if that is the case.
    • However, if it is unclear whether a violation has occurred (e.g., unclear alignment with the workshop’s themes), please complete the review assuming the submission is valid.
    • If grounds for desk rejection are identified, a brief explanation should be provided.
  • A 1–2 paragraph critical review covering the paper’s overall soundness, clarity, and impact potential (see above for details).
  • A recommendation to accept or reject the submission.
  • A checkbox to recommend outstanding submissions for a contributed talk at the workshop. If selected, please specify the relevant contributed talk track (details available here) and provide your justification for the recommendation.

Please note there will be no discussion period; reviews and decisions will be final.

Questions?

Contact us at constrainedml@gmail.com.