joker-zxr/my-skills/aistats-workflow

AISTATS Workflow

Use this as the project-management skill for an AISTATS submission. Replace all dates with the current official timetable and work backwards from OpenReview cutoffs.

AISTATS is a conference, not a journal: it has no standing editor-in-chief and no article-processing charge. The rotating leadership is the per-edition General Chairs and Program Chairs (2026 General Chairs: Emtiyaz Khan and Yingzhen Li; Program Chairs: Arno Solin and Aaditya Ramdas, verified 2026-06-22), and the cost model is registration fees, not APCs — PMLR proceedings are open-access with no author fee. Conference organizers rotate yearly, so re-check the current CFP and organization page rather than carrying a name forward.

Milestones

• Venue fit: confirm the contribution is genuinely at the AI, ML, and statistics interface.
• Evidence lock: freeze theorem statements, assumptions, experiments, simulations,

baselines, reproducibility checklist, and supplement contents.

• Abstract deadline: submit real title, abstract, authors, subject areas, and conflicts.
• Full-paper deadline: upload anonymous PDF and required OpenReview fields.
• Supplement deadline: upload anonymized appendix/code/data if current cycle allows or

requires separate supplementary material.

• Review release: triage correctness, statistical validity, novelty, clarity, and

reproducibility concerns.

• Discussion period: draft concise anonymous text-only clarifications; avoid forbidden links

or unsupported new results.

• Decision: archive reviews, discussion, meta-review, decision, and submitted versions.
• Acceptance: prepare PMLR camera-ready files, registration, in-person presentation, and

public artifact release.

Backward plan from the paper deadline

These offsets are planning heuristics only — anchor every one of them to the current official timetable, never to a previous cycle's calendar.

Failure modes by stage

• Theory still moving at week 3 forces last-minute theorem edits nobody has verified — the

classic AISTATS correctness reject in the making.

• Leaving format conversion to the final week surfaces overflow late, because two-column math

is far tighter than a one-column working draft.

• Skipping the mock review forfeits the only chance to hear "your experiments violate

Assumption 2" from a friend instead of a reviewer.

• Building the supplement after the paper deadline under time pressure is how anonymity

leaks ship.

Coordination notes

• Assign one named owner for the reproducibility checklist and another for the anonymity

sweep; shared ownership is how both slip.

• Archive the exact submitted PDF, supplement, and checklist, since discussion-phase replies

must quote them precisely.

Output format

[Current stage] idea / experiments / writing / abstract / submission / supplement / discussion / accepted
[Next official deadline] <date and source, or unknown>
[Critical path] <three tasks that determine readiness>
[Risk register] <page/anonymity/statistical evidence/supplement/reproducibility/presentation>
[Owner map] <task -> person or role>