Itcs Artifact Evaluation

ITCS Artifact Evaluation

ITCS has no artifact-evaluation track. There are no ACM/IEEE artifact badges, no Available/Functional/Reusable/Reproduced tiers, no code submission, and no evaluation committee inspecting a package. It is a pure-theory venue: the thing that must be independently verifiable is the mathematics. This skill exists because authors arriving from empirical CS (where "artifact evaluation" is a real post-acceptance track) need to know what plays that role at ITCS — and what does not apply.

The ITCS "artifact" is the proof

The artifact that a reviewer inspects is the argument itself. Making it "evaluatable" means:

• Complete proofs of every central claim, inside the submitted PDF (body sketch + appendix full proof). This is the single most important "artifact" obligation and it is enforced at submission, not post-acceptance (see itcs-submission and itcs-reproducibility).
• Self-contained definitions so the claims can be verified without your other papers open.
• Pinned dependencies — every borrowed theorem cited to a precise statement, so a reviewer can audit what you assume versus prove.

There is nothing to "install," nothing to "run," and no badge to earn. The equivalent of a reproduced result is a reviewer following the proof to the end and agreeing.

The public full version as the durable, verifiable record

The closest thing ITCS has to an "available artifact" is the full version on arXiv / ECCC / IACR ePrint, which authors are encouraged to post under lightweight double-blind:

• It is the permanent open record of the complete argument — the analogue of an archived, DOI-bearing artifact, except it is the paper itself, open access, mirrored by LIPIcs on acceptance.
• It does not substitute for proofs in the submitted PDF. During review the PC judges the submission; a central proof that lives only in the preprint counts as absent (see itcs-supplementary).
• Keep the preprint and submission in sync so the "artifact" a later reader verifies matches the one the PC accepted.

When a paper does have computational content

Occasionally an ITCS paper leans on a small computation — a search that found a gadget, a solver certifying a finite separation. Make that checkable, in the spirit of an artifact but without a track to submit it to:

• Include the finite object (the graph, code, certificate) so verification is a static check, not a re-run. A reviewer confirming that a supplied 12-vertex graph has the claimed property is the theory analogue of "reproduced."
• State the search space and the tool/version so the computation is auditable in principle.
• Do not host it on a personal repository during review — that is a lightweight-double-blind leak; fold small objects into an appendix (see itcs-experiments).

What does NOT transfer from empirical-CS artifact tracks

Verifiability checklist (the ITCS "artifact" pass)

[Proofs-in-PDF]   every central claim fully proved inside the submitted PDF? yes/no
[Self-contained]  claims verifiable without the authors' other papers? yes/no
[Dependencies]    borrowed results pinned to precise statements? yes/no
[Full version]    posted (arXiv/ECCC/ePrint) and consistent with the submission? yes/no
[Compute objects] any finite computational witness included and statically checkable? yes/no/NA
[Anonymity]       no personal-repo link exposing identity during review? yes/no

Output format

[ITCS verifiability status] independently-checkable / gaps
[Proof artifact] central claims proved in-PDF? self-contained? dependencies pinned? (yes/no)
[Durable record] full version posted and in sync? yes/no
[Computational witness] finite object included & checkable? yes/no/NA
[Note] there is no badge/track — the deliverable is a checkable proof, not a package
[Fix queue] <ordered; unproved central claims first>

Source: brycewang-stanford/Awesome-Journal-Skills → ITCS-Skills/skills/itcs-artifact-evaluation/SKILL.md