soda-author-response

SODA Author Response

SODA runs a short rebuttal, and short is the operative word. For the 2027 cycle,

initial reviews reach authors by September 1, 2026 and responses are due

September 4, 2026 (checked 2026-07-08 via the SIAM SODA27 submission page,

read through search rendering). SODA 2026's rebuttal phase opened September 8,

2025, in a record-submission year — the window is a house institution now, not an

experiment, but its exact dates, length caps, and mechanics are reset per cycle

(caps for 2027: 待核实 on the HotCRP response form when it opens).

What three days buys — and what it cannot

A SODA response is not a negotiation channel. It exists so the PC can distinguish

"the referee missed Lemma 4.2" from "Lemma 4.2 is actually broken" before the

October decision meeting. Budget accordingly:

| Review problem | Response leverage | Play |

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

| Referee misread a definition or bound | High | Quote the exact line and page; one sentence of correction |

| Referee missed a proof in the back matter | High | Point to the section; the full version was submitted, so it is on record |

| "Improvement too incremental" | Low-medium | One paragraph: name the barrier the prior bound sat at and who tried |

| "Why not compare to [paper]?" | Medium | State the relationship in two sentences; concede if the comparison is due |

| Genuine bug found by the referee | Honest disclosure only | Say whether it is patchable; a false "easily fixed" burns the community's memory |

| Taste objection ("not exciting") | Near zero | Skip it; spend words where facts decide |

Preparation happens in August

Three days is enough to assemble a response only if the material already exists.

Between the July submission and September 1:

• Keep a claims ledger: every theorem, its exact statement, and the page where

each ingredient is proved. Misreading corrections then take minutes, not hours.

• Pre-draft the two paragraphs you can predict: the significance defense (which

prior bound, which barrier) and the closest-related-paper comparison.

• Ensure all coauthors are reachable in the September 1-4 window across time

zones. A rebuttal blocked on a sleeping coauthor is a self-inflicted loss.

Writing discipline

• Numbered points, referee by referee. PC members skim; a wall of prose loses

the one correction that mattered.

• Anchor every reply to a line in the submitted PDF. "See Section 5.3,

Lemma 5.7, page 31" is the strongest sentence a theory rebuttal can contain.

• No new theorems. A result proved after the deadline is not part of the

submission; at most, note that a referee's suggested extension appears to follow

and will be included in the full version.

• Stay anonymous. The lightweight double-blind rules still apply: no names, no

"in our previous paper," no links to identifiable repositories.

• Concede cleanly. One honest "the referee is right; the fix is X and costs a

constant factor" earns more trust than three defensive paragraphs.

Response skeleton

We thank the referees. We address the substantive points.

Reviewer A
A1 (main theorem "only" improves the exponent): The previous bound of
   O(n^{3/2}) [Ref 12] stood since 2015 and was stated as an open problem
   in [Ref 3]. Our O(n^{4/3}) crosses the barrier identified in [Ref 12, §6].
A2 (correctness of Lemma 4.2): The referee reads d as the maximum degree;
   it is the average degree (defined p.7, l.4). The claimed counterexample
   has average degree 3 and satisfies the lemma.

Reviewer B
B1 (missing comparison to [X]): [X] solves the unweighted case; our
   Section 2.1 reduction does not preserve weights. We will state this
   explicitly in Section 1.
B2 (typo in recurrence, p.19): Correct; the exponent should read k-1.
   The solution of the recurrence and all downstream bounds are unchanged.

The 72-hour schedule that actually works

Multi-author rebuttals fail on coordination, not content. A schedule that has

survived real September windows:

| Hour (from review release) | Action | Owner |

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

| 0-4 | Full read; classify each point F/L/S/C (soda-review-process) | Lead author |

| 4-12 | Claims-ledger lookups for every F and C point | Whole group, split by section |

| 12-24 | First draft, worst review first | Lead author |

| 24-36 | Technical check of every page/lemma reference against the submitted PDF | The coauthor who did blind proof-checking |

| 36-48 | Tone pass: cut defensiveness, cut anything answering an S point at length | Most senior author |

| 48-60 | Freeze; sleep on it | — |

| 60-70 | Final read against the length cap and anonymity rules; submit | Lead author |

| 70-72 | Buffer for HotCRP form surprises | Lead author |

Two structural rules: the person who verifies references must not be the person

who wrote the draft, and the submission happens hours before the deadline, not

minutes — the response form's mechanics (length caps, formatting) are 待核实

until the form opens, and surprises cost exactly the buffer you kept.

After the response

There is no discussion phase with referees. Decisions and final-version

instructions arrive from the PC in October (2027 cycle). If the outcome is

rejection, fold the reviews into the resubmission plan — the natural next

deadlines from a September/October SODA rejection are ITCS (if the contribution is

conceptual), STOC in early November if the fixes are fast, or ESA/ICALP in the

new year (soda-workflow has the calendar).

Output format

[Response plan] <points to answer, points to skip, concessions>
[Evidence anchors] <PDF page/lemma reference for each reply>
[Risk flags] <any reply that adds unproved claims or breaks anonymity>
[Coauthor sign-off] <who reviews the draft before the September deadline>