brycewang-stanford WSDM Workflow
Plan a WSDM campaign around the venue's defining rhythm: one deadline a year, in August, for a conference the following February - with no rebuttal stage between them. Missing August costs twelve months, and nothing after August can improve the outcome. Every date below is a 2026-cycle anchor (abstract 2025-08-07, paper 2025-08-14, notification 2025-10-23, conference 2026-02-22 to 26, Boise); the 2027 edition (20th, Hong Kong, February 15-19, 2027) had not published verified deadlines at this pack's check date - confirm them at wsdm-conference.org/2027 before committing the plan.
The annual clock
T-26w fit decision (wsdm-topic-selection) + evidence plan
T-20w datasets locked, split manifests written, baselines running
T-12w main results stable; ablation matrix complete
T-8w full draft; adversarial read #1 -> objection inventory
T-4w artifact repo review-ready; ethics section drafted
T-1w abstract registered (bidding-quality title+abstract)
Aug T=0 paper deadline, AoE <- everything ends here
Oct ~T+10w notification <- no interaction in between
Nov+ camera-ready + ACM production (dates from acceptance email)
Feb single-track conference (talk/poster, registration, travel)
The unusual constraints, made explicit:
- Front-load quality. Venues with rebuttals let you defend a rushed evaluation; WSDM does not. The T-8w adversarial read is the highest-leverage meeting of the campaign - schedule it like a deadline.
- The August-to-October silence is planning time, not waiting time: prepare the camera-ready artifact conversion and, in parallel, the rejection fork (below), so either outcome executes immediately.
- February logistics are real work: winter travel, visas for the host country (Hong Kong in 2027), industry disclosure approvals for the final version - start each at notification, not in January.
The September pressure valve
Since 2026 the series runs a short-paper track with deadlines about a month after the long-paper deadline (2026 anchor: announced September 11/18, with registry evidence of an extension to September 18/25 - verify the current year's dates directly). Plan uses:
- A long paper that is honestly two results and a wish can be split in July: the strong, focused result targets the short track; the broader project waits a year and arrives with more evidence.
- A near-miss at the long deadline (evaluation incomplete) degrades gracefully to the short track instead of slipping twelve months.
- Do not treat "short" as "unfinished": the 2026 call asked for solid, empirically validated contributions, capped submissions at 10 per author, and desk-rejected placeholder abstracts.
Alternate entrances to the same room
| Entrance | Typical timing (verify per edition) | What it earns |
|---|---|---|
| Long paper | August deadline | Archival full paper + single-track talk |
| Short paper | ~September deadline (since 2026) | Archival concise paper |
| Demonstration | Autumn deadline; 4-page cap in 2026 | Archival demo paper + demo floor |
| WSDM Cup | Task-specific autumn schedule | Challenge standing + workshop-style exposure |
| Doctoral consortium | Autumn deadline | Mentoring + community entry for PhD students |
| Industry Day talk | CFP per edition (2026 theme: LLMs and agentic AI in industry) | Non-archival visibility with the practitioner half of the venue |
A lab shipping a system can rationally take two entrances in one year (e.g., short paper + demo) - check the current calls' dual-submission language first.
Coordinating with the sibling calendar
WSDM's August deadline sits inside a lattice of neighbors (all timings approximate and cycle-volatile - verify each): SIGIR around January/February, TheWebConf around October, CIKM around May, RecSys in spring, KDD's two cycles around August and February. Two standard plays:
- Forward chain: a paper rejected from SIGIR (decisions in spring) has a natural summer revision window before WSDM's August deadline - the reviews arrive in time to act on them.
- Backward chain: a WSDM rejection in October reaches SIGIR or KDD Cycle 2 in winter with fixes applied. Map the chain before submitting, so the October decision triggers execution instead of deliberation.
Never dual-submit along the chain: these are archival venues with standard concurrent-submission prohibitions; the chain is sequential by definition.
Contingencies
- Deadline extension announced: treat extra days as validation time, not new-experiment time; late-added results are the least-reviewed and most-attacked parts of a paper at a no-rebuttal venue.
- A competing preprint appears in July: do not race it with rushed experiments. Add the contrast citation, sharpen the delta sentence, and let the August version be correct rather than first-and-fragile.
- Key experiment breaks at T-2w: invoke the valve - descope the claim to what stands, or pivot to the short track - rather than shipping a table you cannot defend to reviewers you cannot answer.
- Chair announcement contradicts the CFP: newest official instruction
wins; log both with dates (precedence order in
../../resources/external_tools.md).
Team operating rules
- One owner for the EasyChair record: account, abstract text, author order, conflicts, final PDF upload.
- The objection inventory (started at T-8w) is a living document with named owners per objection; unresolved rows at T-1w force claim reductions.
- The artifact repo freezes at T-1w except README fixes - late code churn breaks the numbers-match-paper invariant more often than it helps.
- Log every date-fact with its source URL the day you verify it; August deadline drift (extensions happen) is checked weekly in the final month against the official site, not social media.
Output format
[Cycle] target edition + verified deadlines (source + check date)
[Milestones] T-26w..T-0 with owners; current slack per milestone
[Valve decision] long / long+short split / short / defer, with reason
[Alternate entrances] pursued this year: <list or none>
[Chain map] pre-WSDM source venue / post-WSDM destination, with dates
[Logistics] visa / registration / disclosure lead items and start dates