brycewang-stanford ITCS Related Work
At ITCS, related work carries a heavier burden than at a depth venue: since conceptual novelty is the selection criterion, your positioning must prove the contribution is genuinely new, not merely stronger. A reviewer's most damaging finding is not "someone proved a better bound" — it is "someone already asked this question" or "this is a known model under a new name." The related-work discussion is where you win or lose the novelty claim, so it is a load-bearing part of an ITCS paper, not an obligatory appendix.
Delta-first positioning
Lead each comparison with the delta, not the citation:
- Name the closest prior model or question, state what it did, and then state precisely what it did not ask that you do. "Prior work bounds the prover's computation; none bounds its memory of the interaction" is a delta. "Many works study interactive proofs [3-19]" is a dump.
- Distinguish new question from new answer. If the question existed and you answer it
better, say so honestly — and reconsider whether ITCS is the right venue (see
itcs-topic-selection); if the question is new, make that the headline of the paragraph. - Address independent and concurrent work plainly. In a small community with heavy arXiv/ECCC posting, a near-simultaneous related idea is common; acknowledging it and drawing the distinction reads as strength, not weakness.
Cover the right lanes
An ITCS paper usually sits at a seam between areas, so its related work must reach across lanes a single-area reader might miss:
| Lane | Typical venues to survey | Why it matters at ITCS |
|---|---|---|
| Your core area's depth work | STOC, FOCS, the specialist venue (CCC/SODA/TCC) | Shows you know the hard results your idea sits near |
| The other area you connect to | wherever that field publishes | An ITCS bridge is only novel if both banks are cited |
| Prior models/definitions near yours | ITCS/ICS back-catalog, journals | This is where "already asked" hides — search it hardest |
| Folklore and surveys | monographs, lecture notes, Dagstuhl reports | New questions are often "known folklore"; pre-empt that charge |
Scan the ITCS/ICS back-catalog (dblp conf/innovations, LIPIcs volumes) specifically: because
ITCS is where new models are introduced, the venue most likely to already contain your idea is
ITCS itself.
Honesty under lightweight double-blind
ITCS review is lightweight double-blind: the PDF hides author names, affiliations, and emails, but the reference list is NOT anonymized, and authors are encouraged to post a non-anonymous full version to arXiv/ECCC/ePrint. Practical consequences for related work:
- Cite your own prior work normally in the third person. Do not anonymize references or write "[removed for anonymity]." Refer to your own earlier papers as "[Author, YEAR] shows..." the way you would cite anyone — the reference stays in the list, the prose stays neutral.
- Do not write "in our previous work [12] we..." — that first-person self-reference is the
classic leak the anonymity sweep must catch (see
itcs-submission). - Do not sabotage coverage to hide identity. Omitting the obviously-relevant paper by your own group to avoid a tell is worse than the tell: reviewers notice the gap, and it damages the novelty case. Cite it neutrally.
Positioning the novelty claim itself
Make the newness checkable, the way the rest of the paper makes theorems checkable:
- State the novelty claim explicitly ("to our knowledge, no prior work models X as Y") rather than implying it — an explicit claim is one a reviewer can confirm or refute, which is the honest posture.
- Bound the claim. "New, to our knowledge, in the unconditional setting" is safer and more credible than a blanket "first ever."
- Credit the lineage. Showing your idea descends from a recognizable line ("this extends the non-malleability philosophy to ...") makes it legible and generous without weakening the delta.
Common failure modes
| Symptom | Why it fails at ITCS | Fix |
|---|---|---|
| Citation dump with no deltas | Does not prove novelty, the selection axis | Lead each comparison with what prior work did not ask |
| Missed the closest prior model | "Already asked" — fatal to a novelty paper | Search the ITCS/ICS back-catalog hardest |
| Anonymized/removed references | Violates the not-anonymized reference rule | Cite normally in the third person |
| "In our prior work [12] we..." | Anonymity leak under lightweight double-blind | Rewrite as a neutral third-person citation |
| Blanket "first ever" claim | One counterexample sinks it | Bound the claim to a precise setting |
| Only your own area cited | An ITCS bridge needs both banks | Survey the connected area too |
Output format
[ITCS positioning status] novelty-proven / at-risk / unclear
[Closest prior model] <named> ; delta: <what they did NOT ask that you do>
[Back-catalog check] searched ITCS/ICS + connected area for a prior version? yes/no
[Novelty claim] stated explicitly and bounded to a setting? yes/no
[Anonymity] references un-anonymized; no first-person self-cites? yes/no
[Fix queue] <ordered edits>