FOCS submissions carry no length cap, and the CFP encourages submitting the full version of the paper — but everything outside the abstract, the references, and the first ten pages is read only at the committee's discretion (FOCS 2026 CFP, checked 2026-07-08). Style at this venue is therefore an exercise in front-loading: the opening ten pages must let a committee member from a different corner of TCS decide the paper's fate fairly, even if they never turn to page eleven.
A budget that survives contact with real papers:
| Pages | Content | Failure mode it prevents |
|---|---|---|
| 1–1.5 | Problem, headline theorem(s) informally, why they matter | Reviewer cannot state your result after page 2 |
| 1.5–3 | Context: prior bounds, the obstacle, your delta stated precisely | "Novelty unclear" from a non-specialist |
| 3–6 | Technical overview: the proof's actual ideas, with real definitions | "Contribution may be incremental" — ideas were invisible |
| 6–9.5 | Formal statements of all results; proof of one representative lemma in full | Committee cannot check that formality matches the informal claims |
| 9.5–10 | Organization note pointing into the body | Diligent reviewer gets lost after the window |
Every theorem you want credit for must be stated by page ten. A result whose first appearance is page 23 does not exist for most of the committee.
State each headline result twice: once in the introduction in plain language with all quantifiers honest, once formally with every hypothesis. Keep the two mechanically linked so they cannot drift apart across revisions:
% preamble: restatable theorems, one source of truth
\usepackage{thmtools, thm-restate}
\declaretheorem[name=Theorem]{theorem}
% introduction (inside the ten-page window)
\begin{restatable}[Main]{theorem}{mainthm}\label{thm:main}
For every $\varepsilon>0$ there is a deterministic
$O(m^{1+\varepsilon})$-time algorithm that ...
\end{restatable}
% body or appendix, before the proof
\mainthm*
\begin{proof} ... \end{proof}
The thm-restate discipline is the cheapest insurance against the classic
FOCS embarrassment: an introduction promising a bound the formal theorem
quietly weakens.
FOCS has used double-blind review for several cycles (2023–2026 CFPs all require it), with the stated purpose of enabling an unbiased first judgment — not making authorship undiscoverable. The prose consequences:
The abstract is guaranteed reading and often the only text a non-assigned PC member sees before the discussion. A FOCS abstract that does its job contains, in order: the problem in one sentence; the main bound or theorem with honest quantifiers; the one-line comparison to the previous best; the technique named in a phrase; and, when the paper has them, the downstream consequences. What it omits: motivation boilerplate ("X is a fundamental problem"), citations, and any claim the formal theorems do not support. Write it last, from the restatable statements, so the abstract inherits their honesty.
Single-column pages at 11 point invite dense macro-heavy notation; committees across subareas punish it. Working rules:
focs-reproducibility
catches drift in theorems, not prose; only discipline catches this).Do the passes in that order — narrative surgery invalidates later audits, so anonymity always goes last.