gbanyan/pdf_signature_extraction
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
96 Commits 4 Branches 1 Tag
paper-a-v4-big4
Commit Graph

11 Commits

Author SHA1 Message Date
gbanyan 1e8466f7a8 Paper A v4.3: unify Firm A positioning to out-of-sample templated-end target 
Finishes the BCD re-anchor chassis (audit critique #1): Firm A was
inconsistently framed as both a "within-Big-4 case study" and an
"out-of-sample target". Harmonised to a single label, "out-of-sample
templated-end target" (held out of the calibration negative anchor;
scored against the normative Firms-B/C/D baseline), across:
- §I contribution #3 (title + body)
- §III-H.2 (opening trio BCD/Firm-A/non-Big-4; sub-header; role sentence)
- §V-C body (removed the dual case-study/out-of-sample phrasing)
(§V-C header already fixed in ac3372d.)

Zero "case study" wording remains; no numbers changed. codex gpt-5.5
focused check: all consistency items PASS, no new findings.

Also restore the BCD+non-Big-4 joint ICCR Wilson CI [0.000001, 0.000015]
to the §IV-M Table XXI note (three-scope CI symmetry; the one MINOR
completeness gap surfaced by a codex old-vs-new content diff, which
otherwise confirmed no substantive content was dropped by the trim).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-05 02:11:09 +08:00
gbanyan ac3372d2d2 Paper A v4.3: restructure §III baseline-first + deep trim (audit #3/#4) 
#3 Reorder §III around "establish normative baseline → show who deviates":
new order A–O with I=Normative Baseline + inter-CPA coincidence floor
(old L.0–L.3), J=Firm-Level Deviation (old L.4+L.6), K=Why the
distribution gives no threshold (old §I distributional + L.5), L=K=3
partition (old §J), M=Convergent checks (old §K), N=limits (old §M),
O=data source (old §N). ~170 cross-refs remapped (two-pass tokenized),
incl. 5 spelled-out "Section III-X" refs.

#4 Deep trim: §III 10,960→8,461w (−23%). Removed §III↔§IV-M and
§III↔§IV-F table duplication (Results keeps canonical tables; §III keeps
method+headline+pointer); condensed distributional diagnostics;
consolidated repeated caveat. No locked number changed.

Also: §V-C header "Case Study"→"Out-of-Sample Target"; abstract 251→250w;
housekeeping (rm superseded draft_section_L_bcd.md + v4.0 pandoc docx,
remove stale OCR/handoff docs, gitignore .serena/).

codex gpt-5.5 review: 0 BLOCKER / 3 MAJOR / 3 MINOR; 3 MAJOR fixed
(§III-J.2 observed-vs-counterfactual transition, §III-M table pointers
κ→XI/pixel→XIV, §III-N stale tightening figures); 2nd pass clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-05 01:49:09 +08:00
gbanyan 17156516a0 §III-L.0: reword e-signature-adoption rationale as industry background (no interview citation) 
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-04 21:52:05 +08:00
gbanyan 1eb323e959 Paper A v4.2: re-anchor primary calibration to clean BCD 2013-2019 baseline 
- Restrict the calibration negative anchor to Firms B/C/D, fiscal years
  2013-2019 (pre-electronic-signature hand-signing period); B/C/D adopted
  e-signing post-2020 at staggered times, so 2013-2019 is the construct-clean
  baseline. Firm A scored across its full 2013-2023 record against it.
- New locked numbers (codex-audited, Scripts 54/55): per-comparison HC floor
  0.000010; per-signature HC floor 0.0059 [boot 0.0045-0.0073]; per-document
  HC 0.0117 / HC+MC 0.1753; per-firm HC+MC B 0.162 / C 0.225 / D 0.089.
  Firm A observed 0.817 = ~139x the clean floor (was ~70x on all-period BCD);
  Firm A out-of-sample vs clean pool 0.0001 (below floor -> never resembles
  genuine hand-signing). BCD 2020+ robustness: per-sig 0.0105, per-comparison
  0.000036 (~2x pre-2020) quantifies the e-signing contamination.
- Propagated through abstract / Sec. I / III-L / IV-M / V / conclusion;
  0.837 crossover kept corpus-wide; ABCD retained as contamination comparison.
- Grounded the 2013-2019 choice on data (floor drift) + e-sign-adoption
  background, not on in-text interview claims (double-blind).
- Add Scripts 54 (temporal floor stability) and 55 (BCD 2013-2019 primary
  calibration + Firm A scoring).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-04 21:30:06 +08:00
gbanyan 3c7fcc010f Paper A v4.1: BCD-baseline reframe + screening positioning + trim 
- Re-anchor inter-CPA coincidence-rate (ICCR) calibration on a normative
  non-Firm-A baseline (Firms B/C/D); Firm A held out as an out-of-sample
  target. Locked canonical numbers (codex-audited; Scripts 46/52/53):
  per-comparison HC 0.00014->0.000018, per-signature HC 0.0116, per-document
  HC+MC 0.34->0.1905; KDE crossover 0.837 retained corpus-wide.
- Reposition as an operator-tunable, semi-automated screening/triage framework
  (title -> "Automated Screening..."): HC = high-specificity operating point;
  MC band demoted to low-specificity advisory; Firm A = demonstration that the
  screening surfaces a templated end, audit-quality implications deferred.
- Apply codex prose-review fixes: triage-neutral five-way labels, soften
  mechanism/specificity wording, supersede MC claim-strength, update stale
  Appendix script references (40b/43/45 -> 46/52/53).
- Trim pass: compress Sec. V discussion + Sec. III echoes (27.7k -> 26.8k
  words); no substantive content removed.
- Add analysis scripts 45-53 (firm-year trends; BCD-only ICCR recompute;
  canonical-sampler locked numbers; Firm-A out-of-sample; BCD regression +
  cross-firm hit matrix).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-04 19:35:10 +08:00
gbanyan becce857e1 Phase 6 round-7 codex 3-axis review fixes: 11 MAJOR + 5 MINOR 
Codex GPT-5.5 3-axis peer review (paper/codex_review_gpt55_v4_round_3axis.md)
identified 11 MAJOR + 5 MINOR + 0 BLOCKER on three axes: (1) abstract/body
tone consistency, (2) methodology clarity / v3 residue, (3) no implicit
within-CPA or cross-year signature-consistency assumptions. 13 patches
applied across 4 source files; mirrored in paper_a_v4_combined.md.

Axis 1 (tone consistency between abstract and body):
- S I L33: "resolves the ambiguity" -> "provides complementary evidence
  for screening cases where ... hypotheses diverge"
- S I L35: "disproves the distributional-threshold path" -> "does not
  support the distributional-threshold path"
- S I L37 / S V-F L29: "characterise the deployed five-way classifier
  at three units" -> "characterise the deployed HC sub-rule and
  document-level HC+MC alarm derived from the five-way classifier at
  three units" (consistent with S V-H which says only HC sub-rule and
  HC+MC alarm are re-characterised by the present ICCR battery)
- S I L39 / S V-C / S III-L.4: "consistent with firm-specific template,
  stamp, or document-production reuse mechanisms" -> "consistent with --
  but does not independently establish -- firm-level template-like
  reuse, digitisation-pipeline homogeneity, or signing-style
  homogeneity, which descriptor-only data cannot separate (S V-H)"
  (mirrors abstract)

Axis 2 (methodology clarity / v3 residue):
- S III-G: added unit-bridge sentence distinguishing "descriptor-summary
  units" (signature/accountant) from "operational reporting units"
  (per-comparison/per-signature/per-document, S III-L)
- S III-H.2: "The calibration distinguishes two reference populations"
  -> "The supporting diagnostics use two reference populations" with
  explicit "neither is the calibration anchor"
- S III-L.1: "specificity" -> "ICCR refinement"
- S III-L.2: added "descriptive intuition, not an independence
  assumption used for estimation" caveat after the 1-(1-p)^n form

Axis 3 (no implicit signature-consistency assumptions):
- S III-F: hand-signing motivation rewritten as working hypothesis that
  "the classifier does not require ... to hold for all CPAs"
- S III-G A1: added "A1 does not assume temporal stability of
  handwriting or scanning workflow within or across years"
- S III-H.1: added label-caveat paragraph (operational rule outputs,
  not validated ground-truth classes); HC "strong replication evidence"
  -> "image-similarity evidence consistent with replication"; HSC
  "consistent with a CPA who signs very consistently" -> "mechanism not
  resolved by descriptor data alone"; LH explicitly owns that
  cross-year handwriting drift, scanner workflow change, or template
  variant rotation can also yield low max-cosine within a same-CPA pool
- S III-L.6 / S IV-M.6: "same-CPA repeatability signal" -> "observed
  same-CPA-pool excess ... not attributed to within-CPA handwriting
  repeatability"

Deferred (structural, not single-sentence patch): codex S III-I.2 /
S III-J K=2/K=3 deduplication; codex S III-K LOOO / S III-J duplication.
Both are MINOR stylistic redundancies, not reviewer-rejection risks.

DOCX rebuilt via export_v3.py; v4.0_20260515 file refreshed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-16 03:11:53 +08:00
gbanyan 3672c9343e Phase 6 round-6: soften firm-heterogeneity framing, fix DOCX table render 
Framing softening (per partner tone decision: own the limitation rather
than defend the strong claim).

Abstract: "Firm heterogeneity is decisive ... consistent with firm-level
template-like reuse" -> "The framework surfaces pronounced firm-level
heterogeneity ... consistent with firm-level template-like reuse but not
independently diagnostic, since descriptor-only data cannot separate
reuse from digitisation-pipeline or signing-style homogeneity within a
firm; we report it as a scope limitation rather than a mechanism finding."

S V-H Limitations: new bullet "Mechanism attribution for the firm-level
heterogeneity is not identifiable from descriptor-only data." enumerates
three non-mutually-exclusive firm-level mechanisms (template-like reuse /
digitisation-pipeline homogeneity / signing-style homogeneity), notes the
(cosine, dHash) descriptor pair is by construction indifferent to which
mechanism generated a near-identical pair, and lists what additional data
would be needed for attribution.

S VI Conclusion items (3) and (4): "firm heterogeneity quantification" ->
"firm-level heterogeneity surfaced by the framework ... reported as a
framework-discriminative observation rather than a mechanism finding";
item (4) expanded from template/stamp/document-production reuse alone to
the three-mechanism scope, with explicit "not independently establishing"
and S V-H cross-reference.

DOCX export fix (export_v3.py): add missing LaTeX-to-Unicode tokens
(\checkmark, \lvert/\rvert, \lVert/\rVert, \in, \notin, \max, \min,
\log, \ln, \exp, \bullet) that were silently dropping content from
Table III rows 2-4 (integer-jitter robustness check marks empty) and
Table XVIII drift column (|Delta| empty).

Rebuild Paper_A_IEEE_Access_Draft_v3.docx via export_v3.py and install
copy as Paper_A_IEEE_Access_Draft_v4.0_20260515.docx (replaces prior
pandoc-built v4 DOCX which had empty cells in every table header with
LaTeX math and inconsistent column widths). All 43 tables now have
non-empty cells with sub/superscript runs.

Mirrored in paper_a_v4_combined.md for consistency with the single-file
combined source.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-16 02:16:05 +08:00
gbanyan 4efbb7f4b8 Phase 6 round-5 codex-review fixes: minor + nit cleanup 
Codex round-4 verification (commit bbb662a) disposition: Minor revision.
All prior blockers and majors confirmed resolved. Three small items remained.

MINOR (1 of 2 addressed in markdown source):
- Appendix rendered AFTER References (combined L1132 vs L1227), but IEEE
  convention places appendices BEFORE references. Swapped concatenation
  order in combined-file regeneration: abstract -> intro -> related_work
  -> methodology -> results -> discussion -> conclusion -> APPENDIX ->
  REFERENCES -> declarations -> impact. Combined file now has Appendix A
  at L1132 and References at L1194.

MINOR (deferred to typesetting):
- Table A.II is prose-heavy for IEEE double-column layout. This is a
  table-formatting concern for the LaTeX/DOCX export step (table*, smaller
  font, or column-break adjustments), not a markdown-source issue.
  Documenting as a known typesetting consideration for the export pipeline.

NIT:
- Table A.II referenced "§IV-M.4 footnote" but the content at §IV-M.4
  L1007 is inline prose, not a footnote. Changed to "(§IV-M.4)".

Artefacts:
- Combined manuscript regenerated: paper_a_v4_combined.md, 1316 lines.
- Appendix A.1 (BD/McCrary) + A.2 (Diagnostic Summary) precede References.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-15 20:05:24 +08:00
gbanyan bbb662acd1 Phase 6 round-4 codex-review fixes: blocker + 2 majors + minors 
Codex round-3 verification (commit 9e68f2e) flagged remaining items:

BLOCKER:
- Appendix A Table A.I was inside an HTML comment but visible prose at L1268
  and L1275 referenced it as if it rendered. Un-commented Table A.I (same fix
  pattern as Tables I-IV in round-3).

MAJOR:
- Table XXVII (diagnostic summary) appeared in §III-M at L593 before Tables
  III-XXVI in §IV, breaking sequential IEEE-style numbering. Moved the table
  to Appendix A as Table A.II (new §A.2 "Diagnostic Summary"). §III-M now
  references "Appendix A Table A.II" instead of hosting the table inline;
  §VI Conclusion contribution (8) updated similarly. Appendix A heading
  generalised to "Supplementary Diagnostic Detail" with §A.1 (BD/McCrary)
  and §A.2 (Diagnostic Summary).
- §III-M L614 rate-definition conflation: the sentence "per-signature and
  per-document rates ($0.11$ and $0.34$ respectively under the deployed
  any-pair HC + MC alarm)" mixed unit labels. Rewrote to label each rate by
  its actual unit and source table: per-signature any-pair HC ICCR ($0.11$;
  Table XXII) and per-document HC+MC alarm-rate ICCR ($0.34$; Table XXIII).

MINORS:
- L400 §III-L stale ref retargeted: "operational signature-level classifier
  of §III-L" -> "of §III-H.1 calibrated in §III-L".
- L663 §III-L stale ref retargeted: "KDE crossover used in per-document
  classification (Section III-L)" -> "(Section III-H.1)".
- L1146 Conclusion "corpus-universal" overstated generality -> "across the
  tested eligible scopes" (non-Big-4 evidence is cosine-axis only, not full
  calibration scope).
- L1024 Results §IV-M.4 footnote: np.argmax / np.unique implementation
  detail moved to "alphabetically-ordered tie-break ... full implementation
  detail in the supplementary materials" (less script-internal prose).

Artefacts:
- Combined manuscript regenerated: paper_a_v4_combined.md, 1316 lines.
- Final main-text table sequence: I, II, III, ..., XXVI (26 tables, all
  sequential in rendered order). Appendix tables: A.I, A.II.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-15 19:59:56 +08:00
gbanyan 9e68f2e1d3 Phase 6 round-3 codex-review fixes: blockers + majors + minors 
Resolved Codex review (gpt-5.5 xhigh) findings against b6913d2.

BLOCKERS:
- Appendix B reference mismatch: rewrote all main-text "Appendix B" references
  to "supplementary materials" since Appendix B is now a redirect stub. Affected
  the SSIM design-argument pointer, threshold provenance, byte-level
  decomposition, MC band capture-rate, and backbone-ablation table references
  across §III-F / §III-H.1 / §III-H.2 / §III-K / §III-L.4 / §III-M / §IV-F /
  §IV-J / §IV-K / §IV-L / §V-C / §V-H.
- Table rendering: un-commented Tables I-IV (Dataset Summary, YOLO Detection,
  Extraction Results, Cosine Distribution Statistics) which were inside HTML
  comment blocks and would not have rendered in the submission.
- Table numbering out of order: Table XIX appeared before Tables XVI-XVIII.
  Renumbered XIX -> XVI (document-level worst-case counts), XVI -> XVII (Firm x
  K=3 cross-tab), XVII -> XVIII (K=3 component comparison), XVIII -> XIX
  (Spearman correlation). Cross-references updated in §IV-J / §IV-K and §V-C.
- Table V mis-citation: §IV-C said "KDE crossover ... (Table V)" but Table V is
  the dip test. Dropped the (Table V) tag; crossover is a textual finding.
- Submission cleanup: wrapped the archived Impact Statement section heading and
  body inside the existing HTML comment (was rendering). Funding placeholder
  wrapped in HTML comment with a TO-DO note (won't render but is preserved as
  reminder).

MAJORS:
- Line 1077 numerical conflation: rewrote the §V-C / §III-L.4 paragraph that
  labelled Firm A's per-document HC+MC inter-CPA proxy ICCR of 0.6201 as a rate
  "on real same-CPA pools." 0.6201 is a counterfactual proxy under inter-CPA
  candidate-pool replacement, not the observed rate. Added explicit disambig:
  the corresponding observed rate from Table XVI (formerly XIX) is 97.5%
  HC+MC for Firm A; the proxy and observed rates measure different quantities.
- Residual "validation" language softened: "Dual-descriptor verification" ->
  "Dual-descriptor similarity"; "we validate the backbone choice" -> "we
  support the backbone choice"; "pixel-identity validation" -> "pixel-identity
  positive-anchor check"; "## M. Validation Strategy and Limitations under
  Unsupervised Setting" -> "## M. Unsupervised Diagnostic Strategy and Limits".
- "Specificity behaviour" overclaim: "characterises the cosine threshold's
  specificity behaviour" -> "specificity-proxy behaviour" (methodology §III-L.0
  and discussion §V-F).
- "Prior published / prior calibration" ambiguity: replaced "prior published
  per-comparison rate" with "the corpus-wide rate reported in §IV-I"; replaced
  "(prior published operating point)" with "(alternative operating point from
  supplementary calibration evidence)" in Tables XXI; replaced "prior reporting
  and the existing literature" with "the existing literature and the
  supplementary calibration evidence."

MINORS:
- Line 116 Bayes-optimal qualifier: "the local density minimum ... is the
  Bayes-optimal decision boundary under equal priors" -> "In idealized
  two-class mixture settings with equal priors and equal misclassification
  costs, the local density minimum ... coincides with the Bayes-optimal
  decision boundary."
- Stale section refs: §V-G for the fine-tuning caveat retargeted to §V-H
  Engineering-level caveats (where it lives after the §V-H reorganisation);
  §III-L for the worst-case rule retargeted to §III-H.1; "Section IV-D.2"
  (nonexistent) retargeted to "Section IV-D Table VI."
- Abstract / Introduction "after pool-size adjustment": separated the
  document-level D2 proxy ICCR claim from the per-signature logistic regression
  claim. Now: "Per-document D2 inter-CPA proxy ICCRs differ by an order of
  magnitude across firms ... a per-signature logistic regression confirms the
  firm gap persists after pool-size control."

NIT:
- Related Work HTML comment "(see paper_a_references_v3.md for full list)"
  -> "(full list in the References section)"; removes the version-coded
  filename reference from the source.

Artefacts:
- Combined manuscript regenerated: paper_a_v4_combined.md, 1312 lines.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-15 18:28:14 +08:00
gbanyan b6913d2f93 Phase 6 round-2 reviewer revisions: §III-H.1 promotion + framing alignment 
Structural:
- Promote operational classifier definition from §III-L.0 to new §III-H.1, so
  the reader meets the five-way HC/MC/HSC/UN/LH rule before the §III-I/J/K
  diagnostic chain instead of ~130 lines after. §III-L renamed to
  "Anchor-Based Threshold Calibration"; §III-L.0 retains only calibration
  methodology, three units of analysis, any-pair semantics, and the FAR
  terminological note. §III-L.7 deleted (redundant with §III-J).
- Reorganise §V-H Limitations into Primary / Secondary / Documented features /
  Engineering groupings (was a flat 14-item list).
- Reframe §III-M from "ten-tool unsupervised-validation collection" to
  "each diagnostic addresses one specific unsupervised failure mode";
  rename "What v4.0 does/does not claim" → "Limits / Scope of the present
  analysis"; retitle Table XXVII.

Framing alignment (cross-section):
- Strip all v3.x / v4.0 / v3.20 / v4-new / inherited lineage labels from
  rendered text (Abstract, Intro, §II, §III, §IV, §V, §VI, Appendix, Impact).
- Replace "Paper A" rule references with "deployed" rule references.
- Soften "validation" to "characterise" / "check" / "screening label" /
  "consistency check" / "support"; "verdict" → "screening label".
- Remove codex-verified spike claims (non-Big-4 jittered dHash, Big-4 pooled
  cosine after firm-mean centring). Only formally scripted evidence
  (Scripts 39b–39e) retained; non-Big-4 evidence framed as corroborating
  raw-axis cosine, not as calibration evidence.
- Strip script-provenance parentheticals from Introduction; defer Script 39c
  internal references and similar to Methodology / Appendix.

Numerical / table fixes:
- §III-C document-count arithmetic: 12 corrupted → 13 corrupted/unreadable,
  verified against sqlite DB and total-pdf/ folder counts (90,282 - 4,198
  no-sig - 13 corrupted = 86,071 → 85,042 with detections → 182,328 sigs →
  168,755 CPA-matched). Table I shows VLM-positive (86,084) and
  processed-for-extraction (86,071) as separate rows.
- Wilson 95% CIs added for joint-rule ICCR rows in Table XXI / methodology
  table ([0.00011, 0.00018] and [0.00008, 0.00014]).
- Unit error fixed: 0.3856 pp / 0.4431 pp → 0.3856 (38.6 pp) / 0.4431 (44.3 pp).

Smaller revisions:
- Pipeline framing: "detecting" → "screening" in Abstract / Intro / Conclusion
  for consistency with the unsupervised-screening positioning.
- "hard ground-truth subset" → "conservative hard-positive subset" throughout.
- §III-F SSIM / pixel-comparison rebuttal compressed from ~15 lines to 4;
  design-level argument deferred to supplementary materials.
- "stakeholders can adopt / can derive thresholds" → "alternative operating
  points can be characterised by inverting" (less prescriptive).
- "the same mechanism extending in milder form to Firms B/C/D" → "similar,
  milder production-related reuse patterns at Firms B/C/D" (mechanism claim
  softened).
- Appendix A "non-hand-signed mode" / "two-mechanism mixture" lineage language
  aligned with v4 framing.

Appendix B:
- Rebuilt as a redirect-only stub. The HTML-commented obsolete table mapping
  (Table IX–XVIII labels with FAR / capture-rate / validation language) is
  removed; replaced with a short paragraph pointing to supplementary
  materials for full table-to-script provenance.

Cross-references:
- All §III-L references for the rule definition retargeted to §III-H.1;
  references for calibration still point to §III-L.
- §III-H references for byte-level Firm A evidence / non-Big-4 reverse anchor
  retargeted to §III-H.2.

Artefacts:
- Combined manuscript regenerated: paper_a_v4_combined.md, 1314 lines
  (was 1346 pre-review).
- Two review handoff documents added:
  paper/review_handoff_abstract_intro_20260515.md
  paper/review_handoff_body_20260515.md

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-15 18:07:31 +08:00