Splash photo of HKU
Conference on Parsimony and Learning (CPAL)
January 2024, HKU

Reviewer Guidelines

Notable Innovations in Our Review Mechanism

CPAL strives for providing every paper with high-quality, accountable reviews, and therefore takes the following actions in addition:

  • Shepherding by an Action PC: Every paper’s final decision, after being recommended by AC, will go through the direct shepherding of all program chairs (led by one “action PC”). The action PC has two main duties:
    • (before final decision released) The action PC will pay particular attention to the borderline cases and the dispute (large score variations) cases, and will be asked to write additional “meta-meta reviews” in those cases and potentially calibrate on top of AC recommendations. Final decisions will be scrutinized and made in a joint meeting by all program chairs.
    • (after receiving the camera-ready) Each accepted paper’s authors will be asked to submit a one-page cover letter, summarizing what revisions are made between the paper’s submitted and camera-ready versions. The action PC will ensure: (1) all “promised” changes by authors during the discussion stage are indeed implemented; (2) no change that is “too substantial” and “unsoliciated” shall be made to the paper, unless in exceptional circumstances where the action PC has to approve case-by-case. The action PC reserves the right to reject a camera-ready submission and exclude it from the conference proceedings.
  • Semi-Open Identity for Accountability (Action PC and/or AC): For every accepted paper, the names of its AC and action PC will be publicly released on its OpenReview page too. For every rejected paper (excluding withdrawals), only the name of its action PC will be displayed. This decision was not reached lightly; but we hope it would meaningfully add credibility and accountability for every paper’s final outcome.

  • Reviewer Rating and “Dynamic Sparse Selection”: each AC will be asked to rate every reviewer in their batch, in terms of timeliness and quality. Program chairs, who know all reviewers’ identities, will compile a list of reviewers sorted by their average ratings received. Reviewers that receive consistent low reviewer rating for multiple papers / from multiple ACs will be excluded from future review processes.

General Guidelines

We all like the Acceptance Criteria made by TMLR https://jmlr.org/tmlr/acceptance-criteria.html and would instruct our ACs and reviewers to honor the same. In particular we note:

“Crucially, it should not be used as a reason to reject work that isn’t considered “significant” or “impactful” because it isn’t achieving a new state-of-the-art on some benchmark. Nor should it form the basis for rejecting work on a method considered not “novel enough”, as novelty of the studied method is not a necessary criteria for acceptance. We explicitly avoid these terms (“significant”, “impactful”, “novel”), and focus instead on the notion of “interest”. If the authors make it clear that there is something to be learned by some researchers in their area from their work, then the criteria of interest is considered satisfied.