aistats-review-process

AISTATS Review Process

Use this to reason about review-stage strategy. Reopen the current CFP, OpenReview group, author instructions, reviewer instructions if posted, and code of conduct before making process claims.

Process model

• AISTATS uses OpenReview for submission and review workflow in recent cycles.
• Reviewers evaluate technical correctness, statistical and machine-learning contribution, empirical support, clarity, reproducibility, and relevance to artificial intelligence and statistics.
• Author discussion is limited. AISTATS 2026 used a discussion period after initial reviews, with text-only author-reviewer discussion and no links.
• Reviewer and author obligations include confidentiality, appropriate conflicts, professional conduct, and respect for anonymity.
• The most useful response is a decision-focused clarification that gives the area chair or meta-reviewer a clean rationale for acceptance or rejection.
• Accepted papers are published in PMLR, so final metadata and camera-ready compliance matter as much as the initial acceptance.

Who reviews here

• The pool mixes ML researchers with statisticians and statistical learning theorists; expect at least one reviewer to read proofs and assumption sets line by line.
• Because AISTATS is smaller and more specialized than NeurIPS or ICML, topical matches are closer, so vague proof sketches get caught rather than skimmed past.
• Borderline theory-plus-experiments papers usually fall on one of three edges: an assumption the experiments do not satisfy, a missing classical-statistics baseline, or a rate claim never checked empirically.

Scoring leverage table