brycewang-stanford FOCS Review Process
FOCS reviewing is run by a program committee assembled per edition under IEEE Computer Society TCMF sponsorship — for FOCS 2026, chaired by Sanjeev Khanna (CFP checked 2026-07-08). There is no standing editorial board and no multi-year continuity of policy: each cycle's CFP is the constitution, and each PC interprets it once. This skill models the pipeline so authors can predict it rather than be surprised by it.
Pipeline anatomy (April → summer, 2026 cycle)
| Stage | Who | What actually happens |
|---|---|---|
| Assignment | PC chair + conflict data | Papers matched to 2–4 PC members using HotCRP topics and your conflict declarations — errors here trace back to sloppy metadata |
| Expert consultation | PC members → subreferees | A PC member covering a distant subarea recruits the specialist your paper needs; assume at least one reader knows the area's folklore |
| Reading under the window rule | Everyone | Breadth readers judge from abstract + references + ten pages; depth readers verify targeted proofs (focs-supplementary) |
| Committee discussion | Full PC (online and/or meeting) | Papers argued comparatively across all of TCS; the champion model dominates — someone must want your paper in |
| Decision | PC | Notification date for 2026: 待核实 (not published at check time; historically midsummer) |
No author-response round appears in the FOCS 2026 CFP, and none is
traditional at this venue (whether any PC-initiated clarification contact
occurs in 2026: 待核实). The practical consequence is architectural: your
submission is your entire side of the conversation (focs-author-response).
What the committee optimizes
A flagship theory PC comparing ~85% of submissions for rejection converges on three questions:
- Is it correct? Verified by depth readers on the proofs that matter; papers with one delicate lemma and a clean overview get that lemma read hard. Unverifiable-as-written proofs draw "cannot confirm the main claim," which functions as rejection regardless of significance.
- Does it move the field? The consequence/obstacle case from
focs-topic-selection, judged by breadth readers who will not decode subarea-internal significance on their own. - Would the program be poorer without it? The comparative question that kills technically-fine papers: a correct, novel, small step loses its slot to a flawed-but-fertile idea more often than authors expect.
Double-blind, the FOCS way
FOCS moved to double-blind years before its ACM sibling — the 2023, 2024, 2025, and 2026 CFPs all mandate it (first double-blind edition: 待核实), with consistent purpose language: enabling an unbiased initial judgment, while accepting that determined identification (e.g., via the arXiv full version) remains possible. Implications for authors:
- Reviewers are instructed not to hold public preprints against anonymity; keep posting to arXiv/ECCC.
- The bar is on your prose and metadata, not on the world's knowledge
(
focs-submissionhas the sweep). - Subreferees also receive anonymized papers; a "who else could have proved this" guess by an expert is tolerated by design, not a policy failure.
How authors can (and cannot) shape the process
The levers are all pre-deadline: topic selections and conflict declarations
steer assignment (focs-submission); the ten-page window determines what
breadth readers know; self-flagged delicate lemmas direct depth readers'
hours to where scrutiny helps you (a checked hard lemma is a champion's
strongest argument). After the deadline there are no levers — no response
round, no supplementary uploads, no revised versions. The corresponding
anti-levers: an overclaimed abstract poisons the committee's trust in
everything else; a proof the depth reader cannot verify converts a champion
into a skeptic; and any attempt to reach reviewers socially is a reportable
breach in a community small enough that it will be recognized.
Interpreting outcomes
ACCEPT -> execute focs-camera-ready; full version public before talk
REJECT, reviews -> triage comments into: correctness findings (fix before
substantive any resubmission), significance findings (usually means
re-aim the framing or the venue), window findings
("couldn't find X" = your p.1-10 failed, not the reviewer)
REJECT, thin -> common at flagship scale; carries almost no information
reviews about the paper's worth — decide the next venue on your
own significance audit, not on two sentences
The venue's rhythm gives every rejected FOCS paper an immediate second life: the STOC deadline arrives in autumn, SODA and ITCS deadlines cluster nearby, and the arXiv version keeps the result claimable throughout. A FOCS rejection delays a strong theory result by months, not years — plan accordingly rather than appealing, since no formal appeals process exists at this venue (any genuine procedural irregularity goes to the PC chair, once, politely).
Reading the reviews you eventually receive
Whether reviews accompany the decision, and in what depth, varies by cycle (2026 practice: 待核实). When they arrive, decode them by reviewer role rather than by score:
- A breadth review that misstates your result is data about your first ten pages, not about the reviewer — the window failed its one job.
- A depth review engaging specific lemmas is the most valuable free
refereeing your paper will ever get; mine it fully before any
resubmission (
focs-author-responsehas the memo protocol). - A short review with high confidence usually means a subreferee's verdict compressed by the PC member who commissioned it; the terseness carries no signal about care taken.
Calibration notes
- Acceptance statistics vary by cycle and are not published in the CFP; do not quote a rate without a per-year source (待核实).
- Award decisions (Best Paper, Machtey Award for best student paper) are made
by the PC from the accepted pool; nothing in the submission process asks
you to apply — eligibility hygiene is covered in
focs-camera-ready.