brycewang-stanford INFOCOM Author Response
Use this after submission, and — more importantly — before it. The distinctive INFOCOM fact is that there is traditionally no author rebuttal: reviewer scores and TPC discussion decide, and you usually never get a turn to reply. So the "author response" that matters most is the one you bake into the PDF. This skill covers the defensive submission, the early-reject checkpoint, the post-reject move, and any actual response window a cycle introduces (待核实 each year).
The response that counts: a self-defending submission
Because you likely cannot reply, answer the reviewer in the paper:
- Pre-empt the top objection per subarea. A theory reviewer will question an assumption — state and justify it in the system model. A systems reviewer will call a baseline unfair — tune it and document the budget. A measurement reviewer will doubt generalization — scope the claim.
- State limits before a reviewer finds them. A short "where this does not hold" paragraph disarms the objection you could not otherwise rebut.
- Make every number checkable. Named simulator, seeds, confidence intervals, and (optionally) released scripts turn "I don't believe this" into "I can verify this."
This is the opposite of a rebuttal-driven venue, where authors sometimes ship a thin submission planning to "fix it in the response." At INFOCOM that plan fails silently.
Reading an early-reject
INFOCOM's early-reject phase (INFOCOM 2027: 9 Oct 2026) cuts papers before full discussion. Treat it as signal:
[Uniform low scores] the paper did not read as INFOCOM-ready to multiple readers -> structural fix
[Scope mismatch] routed to the wrong subarea, or genuinely off-scope -> re-tag or reroute
[Compliance] double-blind or length violation flagged -> mechanical, fixable next cycle
-> there is no appeal; diagnose which of these it was and act, do not relitigate
An early-reject is rarely a near-miss; a paper the reviewers wanted to discuss survives to the full round. Read it as "this needs real work," not "this was unlucky."
After a final reject
- No rebuttal existed to blame — the decision reflects the paper as written. Extract the named weakness from the reviews and fix that, not the phrasing.
- Reroute by shape. A built-system reject may fit SIGCOMM/NSDI with a deployment story; an analytical reject may fit ToN/JSAC with the full-length treatment the page limit forbade; a focused protocol result may fit ICNP.
- Resubmit deliberately. A lightly edited resubmission to the next INFOCOM meets a similar reviewer pool; change something real (a tuned baseline, a proved case, a larger evaluation).
If a cycle introduces a response window (verify first)
Should a given INFOCOM add an author-feedback/rebuttal period (待核实 — do not assume it), apply the discipline of a scarce turn:
| Pushback | What it signals | Response move |
|---|---|---|
| "The baseline is untuned/unfair" | Soundness doubt | Report the re-run with an equal, documented budget, or justify the choice |
| "The assumption is unrealistic" | Modeling-validity doubt | Bound the gap empirically or scope the theorem; do not hand-wave |
| "Doesn't scale beyond a toy topology" | External-validity limit | Point to the largest evaluated size; add one if the format permits |
| "A proof step is unclear" | Rigor doubt | Give the missing step concisely; point to the appendix/report |
Answer only decision-critical, verifiable points; keep it anonymous; do not add claims the reviewers could not check.
Calibration
- Confirm whether a rebuttal exists this cycle before promising anyone a reply — the default is none.
- The PC discussion decides; even where a response window exists, the revised understanding of your existing paper is what moves a score, not new promises.
- Length and format of any response window vary by cycle; confirm the current instructions.
Output format
[Situation] pre-submission (defensive) / early-reject / final reject / response-window (待核实)
[Defensive audit] <top objection per subarea -> pre-empted in the PDF? yes/no>
[Early-reject diagnosis] uniform-low / scope / compliance -> action
[Reroute plan] <SIGCOMM/NSDI | ICNP | ToN/JSAC | resubmit-with-change>
[If response window] <decision-critical, verifiable points only; anonymity check>