wsdm-author-response

WSDM Author Response

At most ML and data-mining venues this skill would script a rebuttal. WSDM is

different: the venue historically runs no author response phase - a fact

documented as far back as the 2017 reviewing experiment conducted at the

conference itself, and consistent with the short review window (August

submission, October notification in the 2026 cycle). Plan as if you will never

speak to your reviewers. If the current edition's CFP announces a response

phase, that instruction supersedes everything here; check it first.

Preemptive rebuttal: the objection inventory

Since answers must ship inside the PDF, build the rebuttal before submission.

Two weeks out, run an adversarial read with coauthors playing three WSDM

reviewer archetypes - the IR-methods specialist, the applied/industry reader,

and the graph/social-data reader - and log every objection in an inventory:

# objection-inventory.md (one row per anticipated objection)
| # | Likely objection | Archetype | Answer exists? | Where in PDF | Action |
| 1 | Position bias not controlled in log data | IR methods | yes | §4.2 + Tab.3 | none |
| 2 | Baseline X (WSDM'25) missing            | IR methods | no  | -            | add or scope claim |
| 3 | Gains may not survive online serving    | industry   | partial | §5.3     | add latency table |
| 4 | Graph sampled non-uniformly             | graph      | yes | App.A (in-budget) | keep |

Then enforce the rule that makes this venue-specific: every "Answer exists?"

row must resolve to yes + a location or to a claim reduction. "We would

explain in the response" is not an available move at WSDM.

Placement discipline for the answers:

• Objections about validity (leakage, bias, confounds) get answered in the main

evaluation text, not an appendix - reviewers who find the flaw first and the

defense never will score on the flaw.

• Objections about scope get answered by narrowing the claim sentence itself.
• Objections about missing baselines get answered by running the baseline or by

one honest sentence saying why it is out of scope - silence reads as evasion.

What channels do exist after submission

| Channel | Legitimate use | Never use for |

|---|---|---|

| EasyChair note / email to PC chairs | Withdrawal; a factual error you discovered (broken figure, corrupted PDF); a conflict-of-interest problem | Arguing merit, adding results, lobbying |

| Ethics/conduct contact | Reporting reviewer misconduct or a genuine breach | Disagreement dressed as misconduct |

| Camera-ready revision (if accepted) | Implementing reviewer-requested fixes the SPC noted | Adding claims reviewers never vetted |

The bar for contacting chairs is high and the message should be three sentences

or fewer. A team that emails chairs to "clarify a misunderstanding" in a review

is spending reputation at a small, single-track community venue where the same

SPCs return every year.

Reading reviews you cannot answer

Notification arrives with reviews attached and no reply box. Triage them for

the next version, not for argument:

1. Consensus objections (raised by 2+ reviewers) are the revision contract.

They almost always concern evaluation validity or missing comparisons at this

venue - fix them with experiments, not prose.

1. Misreadings are your fault at a no-rebuttal venue: if a competent

specialist misread it, the paper afforded the misreading. Find the sentence

that permitted it and rewrite that sentence in the next version.

1. Taste objections ("not exciting," "incremental") signal venue or framing

mismatch - route through wsdm-topic-selection before resubmitting anywhere.

The rejection fork

WSDM accepts roughly one in six submissions (2025: >100 of >600), so plan the

fork in advance:

• Same work, next WSDM (12 months away): worth it when reviews attack

execution, not fit. Fold in every consensus fix; expect PC overlap - reviewers

remember resubmissions that ignored their reports.

• Sibling venue, sooner: SIGIR (typically January/February deadlines) for

IR-centric work, KDD Cycle 2 (typically February) for general mining, CIKM

(typically May) for a broad CIKM-shaped result, RecSys (spring) for

recommendation, TheWebConf (typically October) for web-platform work. Re-frame

for the destination - a WSDM-tuned paper is not automatically SIGIR-tuned.

• Short paper at the next WSDM: since 2026 the series has a short-paper

track with September deadlines; a focused core finding can land there when the

long version keeps losing on breadth.

Worked micro-example: rewriting an answer into the PDF

Suppose the inventory flags: "an IR reviewer will ask why gains shrink on head

queries." At a rebuttal venue you would draft a reply; at WSDM you edit:

• Draft state: results table shows the shrinkage; text ignores it.
• Bad fix: delete head-query column (reviewers run the slice themselves,

and hiding it converts a weakness into an integrity question).

• WSDM fix: one paragraph after the table: "Gains concentrate on tail

queries (+4.1 nDCG@10) and shrink on head queries (+0.6), consistent with

our mechanism - head queries already receive abundant feedback, so

debiasing has little to correct. This is the intended behavior, not a

failure mode." The objection now has a printed answer with a mechanism

attached, and the SPC can quote it in the meta-review.

The pattern generalizes: locate the evidence that would have gone into the

rebuttal, then write the two sentences of interpretation next to that evidence.

If the current edition adds a response phase

Editions can change process; if the CFP or a chairs' email announces author

responses, adapt rather than assume:

1. Re-read the announced format limits (word caps, scope rules) - do not

import another venue's rebuttal habits.

1. Recycle the objection inventory: rows that made it into the PDF get a

one-line pointer ("§4.3 addresses this"); rows that did not get the

evidence directly.

1. Keep the no-links discipline typical of ACM-venue responses unless the

instructions explicitly allow links.

1. Answer reviewers in their own vocabulary and number the replies against

review points - SPCs skim responses in minutes.

Output format

[WSDM response posture] no-rebuttal cycle confirmed? yes / no (source: current CFP)
[Objection inventory] N objections; unresolved rows: <list or none>
[In-PDF placements] <objection -> section/table map for the top 5>
[Chair-contact check] any legitimate reason to write? yes+draft / no
[Post-decision fork] resubmit WSDM / SIGIR / KDD / CIKM / RecSys / short paper, with re-framing note