Replication package: Weakening in Real Time — The Association Between Artifact Reuse and Citation in Artifact-Dense Research, 2015-2024

Complete data and analysis code for a study of how the association between research artifact reuse (GitHub forks) and citation has changed between 2015 and 2024, across 20,529 arXiv papers carrying an author-designated official code repository. Headline result. The rank association between annual fork flow and annual citation flow weakens by roughly half over the study period. Decomposing the 55 cohort-by-period cells, the variation lies on the calendar-period axis rather than the publication-cohort axis: with period included, the cohort coefficient is -0.0017 (p = 0.74). Period, cohort and age are linearly dependent, so period and age decompositions are observationally equivalent and the archive reports all five specifications rather than selecting one. Bound on the result. The decline is a property of the population, not of its high-impact tail. Enumerating the full sampling frame of 160,150 papers rather than drawing from it, and holding the selection rule fixed, the association among works that are both substantially reused and substantially cited shows no trend: -0.0031 per year (95% CI [-0.0156, +0.0094], n = 3,885 across all ten cohorts), an interval that excludes the full-sample estimate of -0.0241. The test has 80% power against a slope of 0.0151, so this is a null with power behind it rather than an absence of data. Five cohort-window specifications agree. Contents. The stratified sample and the sampling script (which self-verifies against the archived sample and fails rather than silently regenerating a different draw); collected GitHub, Semantic Scholar and OpenAlex records; every analysis script, including the tail enumeration pipeline; and a working memo recording all seven rounds of analysis in order, including hypotheses that were tested and rejected, one intermediate result withdrawn as a pipeline artifact, and the reframing of the estimand from a cohort to a period quantity after an age-period-cohort audit. Snapshot dates. Papers with Code archive mirror 2025-09-10; Semantic Scholar corpus 2026-06-24; GitHub API 2026-07-18 (frame sample) and 2026-07-19 (tail enumeration); OpenAlex 2026-07-18 (frame sample) and 2026-07-19 (tail enumeration). Papers with Code was discontinued during the study period, so the link table cannot be re-collected from the live service; the archived snapshot is included so the sample can be rebuilt exactly. Reproducibility. Analyses from the cell stage onward require no credentials and no network access. All randomness is seeded. See README.md for the pipeline order and a mapping from each manuscript table and figure to the JSON key that produces it.