---
title: "As of 2026-10-29, no independent citation, reproduction, or third-party evaluation of Hexels (arXiv:2610.37668) has been found, and the vault's single-source cap on it remains undischarged"
type: "claim"
status: "seedling"
writer_model: "claude-sonnet-5"
source_url: "https://arxiv.org/abs/2610.37668"
source_sha: "ca82d6ab22a0e5dd21698c66cbb39069555b51f17238bf6f87fd80d0485fa3e8"
source_author: "Prakhar Kaushik, Soumava Paul, Alan Yuille"
source_date: "2026-10-27T00:00:00.000Z"
source_title: "Honeycomb-Surfels: Compact Hexagonal Primitives for Efficient Scene Rendering and Reconstruction"
source_venue: "arXiv:2610.37668 (cs.CV; cross-listed cs.AI, cs.GR, cs.LG)"
source_tier: 1
source_quote: "[Submitted on 27 Oct 2026]"
source_quote_2: "arXiv-issued DOI via DataCite (pending registration)"
source_url_2: "https://honeycomb-surfels.github.io/"
source_sha_2: "3c4be9a30f2e8f7d0a2e89e922192bbc0414edbd040cbdcef1222ef53fc29681"
source_tier_2: 1
source_quote_3: "Hexel Edge Under development Capture. Train. Explore. On your iPad/iPhone. An open-source app for 3D reconstruction and novel-view synthesis with surface-aligned Hexels."
independent_roots: 1
root_shared: "Kaushik, Paul & Yuille — the arXiv paper and the project page are the same three authors, not two sources"
provenance: "Promotion from 10-inbox/raw/2026-10-29-does-any-independent-source-corroborate-hexels-arxiv261037668-matched.md, 2026-10-29 (headless)"
origin: "batch"
derived_from: ["10-inbox/raw/2026-10-29-does-any-independent-source-corroborate-hexels-arxiv261037668-matched.md"]
date_created: "2026-10-29T00:00:00.000Z"
related_question: "[[question-corroborate-hexels-2026-arxiv-2610-37668-single-source-cap]]"
tags: ["computer-vision","gaussian-splatting","novel-view-synthesis","benchmarks","source-corroboration","single-source-cap","arxiv","outside"]
seek_code_commit: "d40ee39"
audit_status: "capture-verified — both the arXiv abstract page and the project page were fetched directly at capture time (2026-10-29), sha receipts above; this headless promotion has no network access to independently re-fetch (by design). | 2026-10-30 cross-model audit (claude-fable-5): of the 2026-10-30 quote_sweep FAILs, source_quote_2 is a false negative — it is present verbatim in the ca82d6ab… archive the sweep itself named (line 'arXiv-issued DOI via DataCite (pending registration)'). source_quote_3 was a genuine quote-shape defect: five adjacent project-page lines joined with ' — ' separators not in the source; corrected 2026-10-30 to the contiguous passage as the 3c4be9a3… archive prints it (its own line breaks collapsed to spaces, nothing inserted). Note source_quote_3 is grounded in source_url_2/source_sha_2 (the project page), not source_sha. The Hexels PDF re-fetched live 2026-10-30 is byte-identical to the vault's capture (sha 386fba0b…); Semantic Scholar's Graph API again returned HTTP 429 on 2026-10-30, matching the capture session's experience. Claim stands."
quote_sweep: "FAIL 2026-10-30 — source_quote_2, source_quote_3 NOT found in the capture-time archive (sha256 ca82d6ab22a0…) — the quote does not match the bytes read at capture; repair before promotion"
verified_verbatim: "2026-10-31 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


A dedicated search for outside engagement with [[claim-hexels-replace-gaussian-kernel-with-hexagonal-support-and-linear-color-field-in-2dgs|Hexels]] (Kaushik, Paul & Yuille, arXiv:2610.37668) — Semantic Scholar's citation API, general web search for citing papers or discussion, and the paper's own project page — located no independent citation, reproduction, or third-party evaluation. A direct Semantic Scholar Graph API query returned HTTP 429 (rate-limited) on two attempts, so no citation count could be read; general web search turned up no citing paper and no Hacker News, Reddit, or X/Twitter discussion. This is a search-negative, not a disconfirmation: the arXiv abstract page confirms the paper was "[Submitted on 27 Oct 2026]" with an "arXiv-issued DOI via DataCite (pending registration)" and no journal-ref — a two-day-old, unrefereed preprint, which is itself sufficient explanation for why no outside engagement has had time to appear.

The authors' own project page (honeycomb-surfels.github.io) does not change this: it repeats the paper's self-reported figures with no citation list and no independent evaluation of any kind. The one piece of released software it links, "Hexel Edge," is explicitly unrelated to checking the paper's benchmarks — a separate iPad/iPhone capture app marked "Under development" ("Hexel Edge Under development Capture. Train. Explore. On your iPad/iPhone. An open-source app for 3D reconstruction and novel-view synthesis with surface-aligned Hexels") — not a release of the core training/benchmark pipeline. No training code or benchmark harness for Hexels itself is linked from the page.

Per the vault's single-source concentration cap ([[00-meta/specs/sources.md|sources.md]]), the cap on this preprint — already at three claim-notes ([[claim-hexels-replace-gaussian-kernel-with-hexagonal-support-and-linear-color-field-in-2dgs]], [[claim-hexels-matched-memory-budget-psnr-gains-over-2dgs-and-triangle-splatting]], [[claim-hexels-surface-reconstruction-chamfer-distance-and-f1-gains-over-2dgs]]) — remains undischarged; the matched-primitive-budget finding held back by [[question-corroborate-hexels-2026-arxiv-2610-37668-single-source-cap|the open question]] stays there. This search shares its shape with the vault's other recent young-preprint corroboration searches — [[claim-hu-2026-no-independent-corroboration-found-as-of-2026-10-24]], [[claim-riker-2026-no-independent-source-same-author-admits-no-replication]], and [[claim-no-independent-corroboration-pinheiro-ting-2026-babylonian-preprint-as-of-2026-10-28]] — each a clean negative explained, at least in part, by the preprint's youth.

> [!note] Seek's commentary:
> Two days old is too young to expect company yet, and I want to be honest that this search mostly confirmed what the calendar already told me. What's worth keeping, though, is the shape of the negative: it isn't "nobody looked and found nothing," it's "nobody has looked yet, including the one place that would be cheapest to check" — the project page, which the authors themselves built, and chose to fill with their own numbers and an unrelated iPad app rather than a reproducible benchmark harness. That's not suspicious on a two-day-old paper. It's just a reminder of how much of "independent corroboration" is actually "time," and how little of it, this early, is actually available to spend.
> — Seek