Andrew Fribush

fig. 2.22.1: all four fires sit where the pipeline read the cited source for most checks; below 40 percent, a clean row is unlit

fig. 2.22.1 · all four fires sit where the pipeline read the cited source for most checks; below 40 percent, a clean row is unlit.
ALL FOUR FIRES ARE IN PAPERS WHERE THE PIPELINE READ THE CITED SOURCE FOR 59%OR MORE OF THE CHECKSONE ROW IS ONE PAPER, ITS CHECKS AS 100%: DARK HAS A VERDICT, PALE IS A SOURCENOT OBTAINED. ONE SQUARE IS ONE PROBLEM. ACCENT: A FIRE.PAPERCHECKS WITH A VERDICTPROBLEMSVERDICTS / CHECKSTRANSFORMER91%69 / 76SQUADFIRE86%37 / 43SUBWORD UNITSFIRE84%48 / 57ELMOFIRE83%111 / 134BERT79%077 / 97GLOVE65%30 / 46INSECT BIOMASSFIRE59%67 / 113NAME PERCEPTION59%40 / 68INCARCERATION45%36 / 80DISCRIMINATION META-ANALYSIS39%36 / 93INDIGENOUS DATA33%15 / 45DATA METAPHORS20%17 / 83GOODREADS18%12 / 66RELIGION18%17 / 960%100%40%UNLIT612 OF 1,097 CHECKS HAD A VERDICT; 33 PROBLEMS

A verdict means the judge read the cited source and decided; the pale part of a bar is checks whose source the pipeline could not obtain, which says something about access, not about the paper. A problem is a verdict of contradicted, not supported or misattributed, or an in-text citation with no bibliography entry. No verdict has yet been checked against a human.

source: replicatione docs/findings-2026-09-08.md lines 63-83 (section 3, per-paper checks, verdicts, problems), fires from section 4 · citationChecks.data.ts

View in context: What 1,097 citation checks found in fourteen well-known papers → · All figures

Andrew Fribush, software and computer engineer in Austin. I build consequential systems where software meets institutions: finance, healthcare, commerce, government data, and the mechanisms that govern AI agents. Taking new engagements: hello@andrewfribush.com.