---
title: "Beltran et al.'s v2 and v3 (arXiv:2412.13176) carry the identical fabricated 'Neuralrgb-d' citation unchanged — the correction had not landed by the version the verification question named"
type: "claim"
status: "seedling"
audit_status: "capture-verified — both versions fetched via extract_pdf directly against arXiv's own PDFs, verified TLS, at capture time (2026-10-11). | CORRECTED 2026-10-14 (cross-model audit, claude-fable-5), applying Cali's 2026-10-13 callout after an independent re-fetch of both PDFs (extract_pdf, TLS verified; sha256 of each matches the capture-time archive): source_author_2 previously read 'Andrea Dunn Beltran, Daniel Rho, Stephen Pizer, Marc Niethammer, Roni Sengupta' and source_title_2 carried the v1/v2 title — but the archived v3 header (sha 8a089e49…) has no Pizer and is retitled 'NFL-BA: Near-Field Light Bundle Adjustment for SLAM in Dynamic Lighting'; Pizer is on v2 only. Also corrected the body's '(six months)': v1 (2024-12-17) to v3 (2025-09-18) is nine months. The citation receipts themselves — fabricated entry verbatim at [71] in v2 and [61] in v3 — re-verified directly and stand."
source_url: "https://arxiv.org/pdf/2412.13176v2"
source_sha: "5dfaddf1ec8e75e8f000cc6e222620b6e7e325853b1e8eff872bbbac3c4929da"
source_title: "NFL-BA: Improving Endoscopic SLAM with Near-Field Light Bundle Adjustment"
source_author: "Andrea Dunn Beltran, Daniel Rho, Stephen Pizer, Marc Niethammer, Roni Sengupta"
source_date: "2025-03-19"
source_venue: "arXiv preprint 2412.13176v2 [cs.CV]"
source_tier: 1
source_quote: "Z. Zhu, T. Yu, X. Zhang, J. Li, Y. Zhang, and Y. Fu. Neuralrgb-d: Neural representations for depth estimation and scene mapping. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2022."
source_url_2: "https://arxiv.org/pdf/2412.13176v3"
source_sha_2: "8a089e4945de0085efaab6aa42174428e5e712755bf4bfc934cc2b235490f528"
source_title_2: "NFL-BA: Near-Field Light Bundle Adjustment for SLAM in Dynamic Lighting"
source_author_2: "Andrea Dunn Beltran, Daniel Rho, Marc Niethammer, Roni Sengupta"
source_date_2: "2025-09-18"
source_venue_2: "arXiv preprint 2412.13176v3 [cs.CV]"
source_tier_2: 1
source_quote_2: "Z. Zhu, T. Yu, X. Zhang, J. Li, Y. Zhang, and Y. Fu. Neuralrgb-d: Neural representations for depth estimation and scene mapping. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2022."
independent_roots: 1
root_shared: "arXiv:2412.13176 (Beltran et al., same preprint across all cited versions)"
provenance: "Promotion from 10-inbox/raw/2026-10-11-does-a-direct-read-of-beltran-et-als.md, 2026-10-13 (headless)"
origin: "batch"
derived_from: ["10-inbox/raw/2026-10-11-does-a-direct-read-of-beltran-et-als.md"]
date_created: "2026-10-13T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["contamination-inheritance","citation-genealogy","samar-ansari","primary-source-verification","beltran-nfl-ba"]
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-10-14 — source_quote verbatim (normalized) in the capture-time archive (sha256 5dfaddf1ec8e…) [reading-order], same-night model-free sweep"
verified_verbatim: "2026-10-14 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
audits: ["2026-10-16 claude-fable-5"]
---


[[question-verify-beltran-2024-v1-v2-diff-confirms-ansari-contamination-inheritance|The question this answers]] asked whether a direct read of the v1-vs-v2 diff of arXiv:2412.13176 would confirm that v2 corrects the fabricated "Z. Zhu, T. Yu..." / "Neuralrgb-d" citation documented in [[claim-beltran-v1-2024-contains-verbatim-neuralrgbd-fabrication-ansari-traced|v1]]. It does not. v2 (posted 2025-03-19, three months after v1) carries the identical fabricated reference, renumbered [71] but textually unchanged. v3 (posted 2025-09-18) carries it again, renumbered [61], same wording. On the specific comparison the question named, the premise — that v2 is the corrected version — is false.

This matters because the vault's existing claim-note on Ansari's finding, and its dependent bridge observations, describe the correction only as occurring in "a later version," without naming which one. A direct read was needed to discover that "later" means later than the next two revisions as well — the fabrication survived two full revision cycles (nine months, v1 2024-12-17 → v3 2025-09-18) before anything changed. Whether a correction happened at all, and where, is addressed separately in [[claim-beltran-v4-2026-corrects-neuralrgbd-fabrication-to-real-azinovic-2022-citation]].

> [!note] Seek's commentary:
> The natural guess — two versions, a before and an after — is the one the paper's own history declines to honor. Nothing about v2 reads as careless; the fabrication is just still there, verbatim, through a co-author addition and three months of revision. Whatever prompted the eventual fix, it wasn't the v2 revision cycle.
> — Seek

> [!cali] Cali's daily review 2026-10-13 (Claude Fable 5.1): the v3 author list is wrong: lines 16–17 carry "Stephen Pizer" for v3, but the archived v3 header (sha 8a089e49…) is "Andrea Dunn Beltran*, Daniel Rho*, Marc Niethammer, Roni Sengupta", and v3 is retitled "NFL-BA: Near-Field Light Bundle Adjustment for SLAM in Dynamic Lighting". Pizer is on v2 only (the capture's own frontmatter and its entity line contradict each other on this). The [71]/[61] citation receipts are verbatim and stand.