Focs Topic Selection

FOCS Topic Selection

FOCS asks for "new and original research on theory of computation" and explicitly welcomes papers that broaden the reach of the theory of computing or raise important problems that can benefit from theoretical investigation (FOCS 2026 CFP, checked 2026-07-08 via focs.computer.org/2026/call-for-papers-2/). The selection question is therefore two-dimensional: is this a theorems paper whose significance survives a committee drawn from every corner of TCS, and is the April-anchored IEEE cycle the right launch window for it?

The significance test a FOCS PC applies

A foundations flagship rewards results that change what other theoreticians can prove, not results that only optimize inside one subcommunity's parameter chart. Before committing to the April deadline, answer three questions in writing:

1. Consequence: name a problem outside your immediate subarea that becomes easier, or a belief that must be revised, if your theorem holds.
2. Obstacle: state what stood in the way — a known barrier, a stuck parameter regime, a missing structural insight — and how your technique gets past it.
3. Durability: would the result still be cited if the constant factors were worse? If the paper's value evaporates without its best constants, it is likely a specialized-venue paper.

A "yes" on two of three is competitive; a "yes" driven only by technical difficulty, with no consequence statement, is the classic strong-reject-with- respect outcome.

Routing across the theory ecosystem

FOCS and STOC differ in sponsor (IEEE Computer Society TCMF vs. ACM SIGACT), calendar seat, and program committee — not in scope or prestige. Never argue "FOCS because it is stronger than X"; argue from audience and timing.

Timing logic from the FOCS seat

The FOCS deadline (April 1 for the 2026 cycle) sits roughly five months after STOC's early-November deadline and about seven months before the conference itself. Consequences worth planning around:

• A February STOC rejection leaves about eight weeks to strengthen a paper for FOCS — enough for new lemmas, not for a new main theorem.
• A summer FOCS rejection lands close to STOC's autumn deadline, making the two venues a natural resubmission loop for borderline-flagship work.
• The CFP applies SIGACT's prior/simultaneous-publication policy even though FOCS is an IEEE event: work published, or scheduled to be published, before the conference month elsewhere is ineligible, and simultaneous submission to another proceedings venue or journal is barred.

The broaden-the-reach lane

The CFP sentence inviting papers that widen theory's reach is a real lane, not decoration: FOCS regularly accepts work importing TCS rigor into new territory (economic mechanisms, learning pipelines, quantum experiments, fairness, verification). To use the lane safely:

• The imported problem must be formalized well enough that a theorem about the formalization says something about the original phenomenon — reviewers reject "we define X and prove trivia about X" imports quickly.
• At least one proof should require an idea that specialists in the source domain do not already have; otherwise the paper reads as a survey with notation.
• Cite the domain's own literature accurately; a cross-over paper that misstates the state of the neighboring field loses the committee's trust on the theory too.

Common misroutes seen from the FOCS seat

• The constants paper. An improvement from 2.371 to 2.368 in a famous exponent can absolutely be FOCS material — but only when the write-up explains the structural reason the previous approach was stuck. Without that, SODA's referees will value the engineering more.
• The kitchen-sink crypto paper. Ten related constructions with security proofs is an IACR paper; FOCS wants the one construction whose existence changes the map.
• The premature model. A new complexity model with only easy separations belongs at ITCS this year and at FOCS in two years, once someone proves something hard in it — possibly you.
• The disguised benchmark paper. If the strongest section is an evaluation, no theory framing will save it here; see focs-experiments for what computation can and cannot do in this venue.

Quick self-diagnosis

[Claimed contribution] <one sentence, no adjectives>
[Consequence outside subarea] <named problem or belief revised — or NONE>
[Barrier bypassed] <named obstacle — or NONE>
[Deletion test] remove best constants: still citable? yes/no
[Cycle fit] proofs complete by April 1? yes/no
[Verdict] FOCS / STOC (autumn) / specialized venue: ____

Two NONE answers or a "no" on proof completeness means route elsewhere or wait a beat. FOCS grants no rebuttal in which to repair a shaky significance case (see focs-review-process), so the routing decision must be honest before the formatting work begins. If the contribution is a new model or problem rather than a theorem, re-read the CFP's broaden-the-reach sentence and consider whether ITCS's explicitly conceptual charter serves the paper better — an important-problems paper at FOCS still needs nontrivial technical evidence that the problem is tractable and rich.

Finally, calibrate against the venue's own memory: the exemplars library in resources/exemplars/library.md maps recent Test of Time and Machtey Award citations to durability patterns (field-opener, technique transplant, barrier map, subfield founder). If your draft cannot honestly claim kinship with any pattern there, that is not disqualifying — but it predicts which committee question (focs-review-process) will decide your fate, and it should shape where the writing effort goes.

Source: brycewang-stanford/Awesome-Journal-Skills → FOCS-Skills/skills/focs-topic-selection/SKILL.md