What does Drift Control for Multi-Window Writing conclude, in numbers, about “A Zero-Hallucination Citation Verifier: Fields, Identifiers, Ledger Matching and Item-by-Item Verdicts”?
Answer: Zero-hallucination verification has three gates. The format gate confirms complete fields, a well-formed DOI, a correct ISBN check digit and a sensible year and volume; it stops copying errors but not a well-made fabrication, because a fabricator can compute the check digit first and assemble a DOI to pattern. The ledger gate matches author, year, title and venue against a local ledger checked record by record against originals; only a title similarity of 0.90 or more can pass, and anything from 0.60 to 0.90 goes to a person. In weighted scoring the "title differs, rest agree" pattern still scores +11.973, so the title must be a gate and not a bonus. Verdicts fall into pass, fix, hold and reject, and only the first two may exist in a print-ready manuscript. In this chapter's fictional sample, F6 with its borrowed DOI and the chimera F5 are the two entries the format checks let straight through and only the ledger stops; and the higher the ledger's coverage, the shorter the hold list and the more suspicious each hold becomes.
Scope of the numbers: Every error rate without a source is an assumed parameter or a measurement on synthetic data; the book is not legal advice and recommends no writing tool or model. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it.
Data table
The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 11.973
| A. variant lev bi word | |
| exact | 1.000 1.000 1.000 |
| case and dash | 1.000 1.000 1.000 |
| one typo | 0.981 0.936 0.750 |
| subtitle added | 0.800 0.821 0.700 |
| reordered | 0.500 0.938 1.000 |
| plausible chimera | 0.732 0.618 0.556 |
| same words, new claim | 0.173 0.917 1.000 |
| unrelated | 0.308 0.164 0.000 |
| B. field m u agree w disagree w | |
| title | 0.950 0.001 +9.892 -4.320 |
| first author surname | 0.970 0.010 +6.600 -5.044 |
| year | 0.900 0.050 +4.170 -3.248 |
| venue | 0.920 0.020 +5.524 -3.615 |
| C. pattern total weight | |
| all four agree +26.185 | |
| year off, rest agree +18.767 | |
| title differs, rest agree +11.973 | |
| only title agrees | -2.015 |
| nothing agrees | -16.228 |
Show the full program
# Chapter 5. p54: Title similarity measures, and Fellegi-Sunter field weights, on FICTIONAL titles.
# Normalise: case-fold, drop punctuation, collapse spaces. Three measures against the ledger title:
# lev = 1 - Levenshtein(a, b) / max(len a, len b) (character edits)
# bi = Jaccard of character bigrams (order-tolerant, typo-tolerant)
# word = Jaccard of word sets (order-blind)
# Part B: agreement weights log2(m / u) and disagreement weights log2((1 - m) / (1 - u)) per field.
# The m and u values are ASSUMED for the illustration, not estimated from data.
import math
import re
LEDGER_TITLE = "Ledger-First Citation Checking for Serial Monographs"
VARIANTS = [
("exact", "Ledger-First Citation Checking for Serial Monographs"),
("case and dash", "ledger first citation checking for serial monographs"),
("one typo", "Ledger-First Citaton Checking for Serial Monographs"),
("subtitle added", "Ledger-First Citation Checking for Serial Monographs: A Field Test"),
("reordered", "Citation Checking for Serial Monographs, Ledger-First"),
("plausible chimera", "Ledger-Based Citation Verification for Serial Monographs"),
("same words, new claim", "Serial Monographs for Ledger-First Citation Checking"),
("unrelated", "A Theory of Shelf Rotation in Small Bookshops"),
]
FIELDS = [("title", 0.95, 0.001), ("first author surname", 0.97, 0.01), ("year", 0.90, 0.05),
("venue", 0.92, 0.02)]
def norm(t):
return re.sub(r"\s+", " ", re.sub(r"[^\w\s]", " ", t.casefold())).strip()
def lev(a, b):
prev = list(range(len(b) + 1))
for i, ca in enumerate(a, 1):
cur = [i]
for j, cb in enumerate(b, 1):
cur.append(min(prev[j] + 1, cur[j - 1] + 1, prev[j - 1] + (ca != cb)))
prev = cur
return prev[-1]
def jac(x, y):
return len(x & y) / len(x | y) if x | y else 1.0
def measures(a, b):
a, b = norm(a), norm(b)
l = 1 - lev(a, b) / max(len(a), len(b))
bi = jac({a[i:i + 2] for i in range(len(a) - 1)}, {b[i:i + 2] for i in range(len(b) - 1)})
return l, bi, jac(set(a.split()), set(b.split()))
def main():
print("A. variant lev bi word")
for name, t in VARIANTS:
l, bi, w = measures(LEDGER_TITLE, t)
print(" {:<22} {:.3f} {:.3f} {:.3f}".format(name, l, bi, w))
print("B. field m u agree w disagree w")
weights = {}
for f, m, u in FIELDS:
weights[f] = (math.log2(m / u), math.log2((1 - m) / (1 - u)))
print(" {:<22} {:.3f} {:.3f} {:>+8.3f} {:>+8.3f}".format(f, m, u, *weights[f]))
patterns = [("all four agree", (1, 1, 1, 1)), ("year off, rest agree", (1, 1, 0, 1)),
("title differs, rest agree", (0, 1, 1, 1)), ("only title agrees", (1, 0, 0, 0)),
("nothing agrees", (0, 0, 0, 0))]
print("C. pattern total weight")
for name, agree in patterns:
tot = math.fsum(weights[f][0] if a else weights[f][1] for (f, _, _), a in zip(FIELDS, agree))
print(" {:<28} {:>+8.3f}".format(name, tot))
if __name__ == "__main__":
main()Single source and citation
Fan, Y. (2026). Drift Control for Parallel Multi-Window Book Writing, Cross-Volume Terminology Graph Audits and Zero-Hallucination Citation Verifiers (Version 0.1.0), Chapter 5: A Zero-Hallucination Citation Verifier: Fields, Identifiers, Ledger Matching and Item-by-Item Verdicts. 5D Library. DOI pending
DOI pending · monograph page goes live after publication · 5DL-B-152 · v0.1.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026152,
author = {Fan, Yafu},
title = {Drift Control for Parallel Multi-Window Book Writing, Cross-Volume Terminology Graph Audits and Zero-Hallucination Citation Verifiers},
year = {2026},
publisher = {5D Library},
edition = {0.1.0},
note = {DOI pending}
}Other chapter answers in this book
- Chapter 1 · Method: Five Failure Modes, Exclusive Locks and the Four Methods of Collation
- Chapter 2 · Concurrency Control: Exclusive-Create Locks, Leases, Version Numbers and Wave Scheduling
- Chapter 3 · The Terminology Graph: Lock Tables, Synonym Drift and Cross-Volume Consistency Audits
- Chapter 4 · Number Parity and Duplicate Sentences: Bilingual Agreement, Program Reconciliation and Near-Duplicate Detection
- Chapter 6 · Style Drift: Gates for Sentence Length, Lexical Diversity, Stock Phrases and Paragraph Openers
- Chapter 8 · Practice: Four Windows Write a Nine-Chapter Book, Five Classes of Injected Error and the Detection Ledger of Four Gates