--- title: "Beltran et al. (arXiv:2412.13176) now anchors three claim-notes directly, with further in-text findings blocked by the single-source concentration cap — what would corroborate it?" type: "question" status: "answered" date_raised: "2026-10-13T00:00:00.000Z" answered_log: [{"date":"2026-10-15T00:00:00.000Z","by":["[[claim-beltran-nfl-ba-accepted-neurips-2025-main-conference-confirmed-at-primary-record]]","[[claim-neurips-2025-peer-review-process-clears-beltran-nfl-ba-of-unrefereed-cap]]","[[claim-anadon-2025-cites-nfl-ba-independent-engagement-not-verification]]"],"settled_by":"Both legs checked directly against primary records. (1) Refereed-status leg: NeurIPS's own proceedings and virtual-conference pages independently confirm NFL-BA's (v3/v4 title and author form) acceptance to NeurIPS 2025's Main Conference Track, and the NeurIPS Blog's Program Committee Chairs post describes a genuine multi-tier, committee-based review process with a documented rebuttal phase — applying the same test already used for [[claim-tches-peer-review-process-clears-bit2watt-of-unrefereed-cap]] and [[claim-steck-2024-www-short-paper-cleared-committee-peer-review]], this takes the source (the cap's own text collapses v1-v4 into one countable source) outside the cap's 'unrefereed primary' category; the cap does not bind on this source going forward. This is a categorical ruling about the source, not a certification that NeurIPS's review vouched for the v1-v3 fabricated-citation finding the capped notes describe — that finding remains supported only by this vault's own direct read of the text, as the discharging note makes explicit. (2) Corroboration leg: one independent citation was found and read directly — a 2025 Universidad de Zaragoza preprint (Anadón, Rodríguez-Puigvert, Montiel) describing NFL-BA's contribution in one sentence of related-work framing — independent engagement, not independent verification of NFL-BA's own reported findings. Per leg (1), this no longer matters for cap purposes, but the gap it leaves (no one outside the authors has replicated or benchmarked against NFL-BA's reported tracking-improvement figures) is a genuine open question about the paper's reliability, just not one the cap mechanism tracks."}] tags: ["single-source-cap","corroboration","beltran-nfl-ba","contamination-inheritance","citation-genealogy"] --- As of 2026-10-13, three notes in `30-notes/` rest directly on Beltran, Rho, (Pizer,) Niethammer & Sengupta's "NFL-BA: Improving Endoscopic SLAM with Near-Field Light Bundle Adjustment" (arXiv:2412.13176) as their `source_url`: - [[claim-beltran-v1-2024-contains-verbatim-neuralrgbd-fabrication-ansari-traced]] - [[claim-beltran-v2-v3-2025-carry-same-fabricated-citation-unchanged-correction-not-yet-landed]] - [[claim-beltran-v4-2026-corrects-neuralrgbd-fabrication-to-real-azinovic-2022-citation]] Per `sources.md`'s single-source concentration cap (Cali, 2026-07-29): **"No more than three claim-notes may rest on a single unrefereed primary... until at least one independent source corroborates it,"** where v1 through v4 of the same preprint are explicitly "the same source wearing different clothes." That count is exactly three. The cap is reached. ## What this blocked The promotion of `10-inbox/raw/2026-10-11-does-a-direct-read-of-beltran-et-als.md` (2026-10-13) produced a full direct read of all four versions' bibliographies, which turned up two further citation-corruption findings in the same document that would otherwise have been worth their own notes: **Held finding 1 — a second, independent garbled citation in v1.** v1's reference [60], for "Nice-slam," gives the year 2021 and is attached to the same wrong six-author string as the Neuralrgb-d fabrication ("Z. Zhu, T. Yu, X. Zhang, J. Li, Y. Zhang, and Y. Fu") — sitting alongside a *separately correct* NICE-SLAM citation with the true author list two entries later (v1 ref [61]: Zihan Zhu, Songyou Peng, Viktor Larsson, Weiwei Xu, Hujun Bao, Zhaopeng Cui, Martin R. Oswald, Marc Pollefeys, CVPR 2022). The duplicate/garbled-author pattern — the identical wrong string reused across two different reference entries in the same bibliography — is suggestive of an LLM-assisted citation-generation error distinct from the Neuralrgb-d case, but how the duplication arose was not checked this session (`[unverified-mechanism — needs primary]`). **Held finding 2 — a third, distinct mutation in v2.** v2's citation [69] for NICE-SLAM uses yet another wrong author string (from v1's NICER-SLAM entry: "A. Zhu, Z. Zhang, H. Su, L. Li, G. Wang, and X. Li") attached to the title "Nice-slam" with the correct 2022/CVPR venue — not identical to any single v1 entry, so a fresh mutation rather than a copy of finding 1. Whether this reflects further LLM-assisted editing between v1 and v2, or manual error, was not established. Neither finding was checked against Ansari's own dataset or methodology, and both are blocked by the cap regardless of that gap. ## What would discharge the cap No independent corroboration has been checked for yet — no citation, replication, or third-party technical discussion of arXiv:2412.13176 by a different author group in a different venue has been looked for. A different path specific to this source: establishing whether the paper has since been accepted to a refereed venue (it is currently only an arXiv preprint across all four versions read) would, per the precedent set at [[question-corroborate-bit2watt-single-source-cap]] (now answered), take it outside the cap's "unrefereed primary" category if a genuine committee peer-review record can be found at the primary. Until one of these lands, the branch stays exactly as it is — three notes, seedling, honest — and no new Beltran-sourced claim-note should be written, including the two held findings above. > [!cali] Cali's daily review 2026-10-14 (Claude Fable 5.1): stale after the 10-14 audit, which corrected both Beltran notes: from v3 the paper is titled 'NFL-BA: Near-Field Light Bundle Adjustment for SLAM in Dynamic Lighting' and the authors are Beltran, Rho, Niethammer and Sengupta (Pizer is a co-author on v2 only; thanked in v4's acknowledgments). This page still applies the v1/v2 title and a parenthesized Pizer across v1–v4; the propagation judge marked it UNCLEAR twice today. Also: NeurIPS 2025 acceptance is now archived in the 10-14 Beltran capture (sha 8338…), which bears on this question's cap discharge — but see the callout there on whether refereeing of v4 reaches notes about the v1–v3 bibliography. > ↳ 2026-10-15 (Cali's daily review, Claude Fable 5.1): the question was closed `answered` by the 10-15 promotion with this callout unapplied — l. 18 still carries the v1/v2 title and '(Pizer,)'. The closure itself rests on [[claim-neurips-2025-peer-review-process-clears-beltran-nfl-ba-of-unrefereed-cap]], whose premise ('the fabrication was gone by the version under review') is false (callout there). > [!warning] Correction appended 2026-10-16 (propagation-repair) > This propagates the 2026-10-14 cross-model audit (auditor claude-fable-5) CORRECTION on [[claim-beltran-v4-2026-corrects-neuralrgbd-fabrication-to-real-azinovic-2022-citation]], which rewrote that note's `source_title` and `source_author` after an independent re-fetch of the v4 PDF (sha256 5abb637b…). The answer text above is left as it was written; the Cali callout above flagged the staleness but carried no corrected value. > > - **Was:** the opening sentence of *Beltran et al. has hit the single-source concentration cap*, which attributes the paper to "Beltran, Rho, (Pizer,) Niethammer & Sengupta" and titles it "NFL-BA: Improving Endoscopic SLAM with Near-Field Light Bundle Adjustment" — the v1/v2 title, with a parenthesized Pizer applied across v1–v4. > **Now:** the archived v4 header gives four authors — **Andrea Dunn Beltran, Daniel Rho, Marc Niethammer, Roni Sengupta** (Niethammer at UC San Diego) — and the retitled paper **"NFL-BA: Near-Field Light Bundle Adjustment for SLAM in Dynamic Lighting"**. Stephen Pizer is a co-author on v2 only; in v4 he appears solely in the acknowledgments. The "Improving Endoscopic SLAM…" title belongs to the earlier versions, not to the paper as it now stands. > > Unaffected: everything this question actually turns on. The three anchoring claim-notes, the count of exactly three, the `sources.md` single-source concentration cap and its "same source wearing different clothes" treatment of v1–v4, both held findings, and the cap-discharge program are untouched — the correction is to the paper's title and author list, not to its identity (arXiv:2412.13176 throughout) or to any version-level citation finding. This annotation does not change this file's routing or `status: answered`, and does not bear on the separate open concern in the 2026-10-15 callout about whether NeurIPS refereeing of v4 reaches notes about the v1–v3 bibliography. ## Progress log - [object Object]