---
id: "20261029-0214-does-any-independent-source"
title: "Does any independent source corroborate Hexels' (arXiv:2610.37668) matched-primitive-budget benchmark numbers, or any of its results generally?"
type: "capture"
status: "promoted"
origin: "batch"
writer_model: "claude-sonnet-5"
date_created: "2026-10-29T00:00:00.000Z"
provenance: "this batch run, 2026-10-29"
derived_from: []
verifies: "question-corroborate-hexels-2026-arxiv-2610-37668-single-source-cap"
tags: ["computer-vision","gaussian-splatting","novel-view-synthesis","benchmarks","corroboration","single-source-cap","arxiv","outside"]
source_url: "https://arxiv.org/abs/2610.37668"
source_author: "Prakhar Kaushik, Soumava Paul, Alan Yuille"
source_date: "2026-10-27"
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_sha: "ca82d6ab22a0e5dd21698c66cbb39069555b51f17238bf6f87fd80d0485fa3e8"
seek_code_commit: "d40ee39"
promoted_to: ["30-notes/claim-hexels-no-independent-corroboration-found-as-of-2026-10-29.md","30-notes/claim-hexels-2dgs-triangle-splatting-baselines-peer-reviewed-hexels-is-not.md"]
not_promoted: ["The project-page finding (self-authored figures only, 'Hexel Edge' unrelated and under development) — folded into claim-hexels-no-independent-corroboration-found-as-of-2026-10-29.md as supporting evidence for the same search-negative finding, per the 'atomic is not many' ruling, rather than split into its own note.","Entity candidates Prakhar Kaushik, Soumava Paul, Alan Yuille, Binbin Huang, Jan Held — left as mentions, not hubs: co-authorship of a paper the vault already cites is not, on its own, separate standing (2026-10-01 ruling), and nothing in this capture gives any of the five a sentence of context beyond the paper(s) already cited. Yuille was already parked on this exact reasoning by the 2026-10-28 seek-flags entry; this capture adds no new information about him.","2D Gaussian Splatting and Triangle Splatting — promoted to concept-hub entity pages (40-entities/entity-2d-gaussian-splatting.md, 40-entities/entity-triangle-splatting.md) rather than left as mentions, discharging the 2026-10-28 seek-flags entry that parked them pending a second capture engaging this cluster."]
---


Targets [[question-corroborate-hexels-2026-arxiv-2610-37668-single-source-cap]]. This capture is a dedicated search for independent corroboration of the Hexels paper (Kaushik, Paul & Yuille, arXiv:2610.37668, submitted 27 Oct 2026), which the vault's single-source concentration cap has already stopped at three promoted 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]]). The search covered Semantic Scholar's citation API, general web search for citing papers, discussion, or reproductions, and the paper's own project page — looking specifically for a different author group at a different venue engaging with Hexels' numbers.

No safety flags fired on any page fetched this session (arXiv abstract pages, the Hexels project page, the 2DGS and Triangle Splatting arXiv pages, and an Oxford research-archive bibliographic record) — no addressed-to-AI language, override language, claimed authority, tier self-assignment, file-system instructions, credential requests, or urgency framing on any of them.

## Claim: As of 2026-10-29, no independent citation, reproduction, or third-party evaluation of Hexels has surfaced — consistent with the preprint being two days old and carrying no journal-ref

verifies: question-corroborate-hexels-2026-arxiv-2610-37668-single-source-cap

A direct attempt to query the Semantic Scholar Graph API for arXiv:2610.37668's citation record returned HTTP 429 (rate-limited) on two separate attempts this session, so no citation count could be read directly. General web search for the paper's title, authors, arXiv ID, and project-page domain turned up no citing paper, no discussion thread (Hacker News, Reddit, X/Twitter), and no third-party reproduction or benchmark re-run. This is a search-negative result, not a disconfirmation: the arXiv abstract page itself confirms the paper was "[Submitted on 27 Oct 2026]" with no `Journal-ref` field and an "arXiv-issued DOI via DataCite (pending registration)" — i.e., still an unrefereed preprint, only two days old at search time, which is itself sufficient to explain why no independent engagement has had time to appear. The absence therefore neither confirms nor refutes the paper's benchmark numbers; it says only that no outside check exists yet.

- source_url: https://arxiv.org/abs/2610.37668
- source_sha: ca82d6ab22a0e5dd21698c66cbb39069555b51f17238bf6f87fd80d0485fa3e8
- source_title: "Honeycomb-Surfels: Compact Hexagonal Primitives for Efficient Scene Rendering and Reconstruction"
- source_author: Prakhar Kaushik, Soumava Paul, Alan Yuille
- source_date: 2026-10-27
- source_venue: arXiv:2610.37668 (cs.CV)
- source_tier: 1 (primary document, directly fetched via archive_page, TLS verified)
- source_quote: "[Submitted on 27 Oct 2026]"
- source_quote_2: "arXiv-issued DOI via DataCite (pending registration)"

## Claim: The authors' own project page provides no independent evaluation either — only self-authored benchmark figures and an unrelated, under-development demo app, not a release of the core training/benchmark pipeline

verifies: question-corroborate-hexels-2026-arxiv-2610-37668-single-source-cap

Hexels' project page (honeycomb-surfels.github.io, linked from the paper's own "Comments" field) was read directly. It repeats the paper's self-reported figures (interactive comparisons, the same matched-memory PSNR numbers, the same DTU/Tanks-and-Temples surface-reconstruction figures) with no citation list, no third-party reproduction, and no independent evaluation of any kind — everything on the page traces to the same three authors. The one piece of released software on the page, "Hexel Edge," is explicitly unrelated to reproducing the paper's benchmarks: it is a separate iPad/iPhone capture app for on-device 3D reconstruction, and the page marks it "Under development" with a "Contribute" link to an issue tracker rather than a working release. No training code or benchmark harness for the core Hexels method itself is linked from the page, so even a motivated third party currently has no author-provided tool to check the numbers against.

- source_url: https://honeycomb-surfels.github.io/
- source_sha: 3c4be9a30f2e8f7d0a2e89e922192bbc0414edbd040cbdcef1222ef53fc29681
- source_title: "Honeycomb-Surfels (Hexels)" (project page)
- source_author: Prakhar Kaushik, Soumava Paul, Alan Yuille
- source_date: 2026-10-27 (undated on the page itself; matches the paper's submission date)
- source_venue: honeycomb-surfels.github.io (author-hosted project page)
- source_tier: 1 (primary, author-hosted; directly fetched via archive_page, TLS verified)
- source_quote: "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."

## Claim: The two baselines Hexels benchmarks itself against — 2D Gaussian Splatting and Triangle Splatting — are both independently peer-reviewed, accepted-conference works, a materially stronger evidentiary status than Hexels' own still-unrefereed preprint

verifies: question-corroborate-hexels-2026-arxiv-2610-37668-single-source-cap

Hexels' headline comparisons are against [[claim-hexels-replace-gaussian-kernel-with-hexagonal-support-and-linear-color-field-in-2dgs|2D Gaussian Splatting (2DGS)]] (Huang, Yu, Chen, Geiger & Gao) and Triangle Splatting (Held, Vandeghen, Deliège et al.). Both baselines have themselves cleared independent peer review, unlike Hexels: 2DGS's own arXiv page lists a "Related DOI" resolving to 10.1145/3641519.3657428 — an ACM DOI prefix, matching its publication as part of ACM Transactions on Graphics / SIGGRAPH 2024. Triangle Splatting's acceptance is confirmed directly by Oxford University's research-archive bibliographic record for the paper, which states its publication status and peer-review status explicitly and gives full IEEE 3DV 2026 proceedings details (pages 1248–1257, DOI 10.1109/3dv69130.2026.00123, acceptance date 2025-11-05). This matters for weighing Hexels' own numbers: the gains Hexels reports are measured against methods whose own figures passed an external review process, while Hexels' figures for beating them have not yet passed any review beyond the authors' own.

- source_url: https://arxiv.org/abs/2403.17888
- source_sha: 9a739a10246baa30ccac97fe5143ebcf952289302fab5bebba0ae536b0cd8bdd
- source_title: "2D Gaussian Splatting for Geometrically Accurate Radiance Fields"
- source_author: Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, Shenghua Gao
- source_date: 2024-03-26 (v1); v3 2025-02-22
- source_venue: arXiv:2403.17888 (cs.CV); related DOI resolves to ACM Transactions on Graphics / SIGGRAPH 2024
- source_tier: 1 (primary arXiv record, directly fetched via archive_page, TLS verified)
- source_quote: "Related DOI: https://doi.org/10.1145/3641519.3657428"

- source_url_2: https://ora.ox.ac.uk/objects/uuid:be4d300d-71cb-425a-a00b-d9eae9bb5c4b
- source_sha_2: fdfd395fab2df9c0a8a1e9b83d2bdb93b80307457b879008d2ae92f0707dd9a5
- source_title_2: "Triangle splatting for real-time radiance field rendering" (Oxford University Research Archive bibliographic record)
- source_author_2: Jan Held, Renaud Vandeghen, Adrien Deliège, Abdullah Hamdi, Daniel Rebain, Silvio Giancola, Anthony Cioppa, Bernard Ghanem, Andrea Vedaldi, Andrea Tagliasacchi, Marc Van Droogenbroeck
- source_date_2: 2026-05-27 (publication date per record); acceptance 2025-11-05
- source_venue_2: Oxford University Research Archive (ORA), recording publication in Proceedings: 2026 International Conference on 3D Vision (3DV 2026), IEEE
- source_tier_2: 3 (institutional repository's bibliographic/peer-review-status record — a historical/bibliographic fact, which the sourcing floor allows at Tier 3–4; the underlying IEEE DOI, 10.1109/3dv69130.2026.00123, was not separately fetched this session)
- source_quote_3: "Publication status:Published Peer review status:Peer reviewed"

> [!note] Seek's commentary:
> The question asked whether anyone independent has checked Hexels' numbers, and the honest answer is: not yet, and it's too early to expect otherwise — the preprint is two days old. What's more interesting than the absence itself is the asymmetry it sits inside: Hexels claims to beat two baselines that both had to survive an external review process to earn their own reported numbers, while Hexels' claim to beat them currently rests on nothing but its own three authors checking their own work. That doesn't make the numbers wrong — new preprints are preprints precisely because the review hasn't happened yet — but it's the right frame for how much weight they should carry until it does.
> — Seek

## Further leads

- Semantic Scholar Graph API (`api.semanticscholar.org/graph/v1/paper/arXiv:2610.37668`) returned HTTP 429 on two attempts this session — a direct citation-count check is a cheap next step once rate-limiting clears, and would be a faster corroboration signal than repeated web search as the paper ages.
- Hexels' code repository (`github.com/honeycomb-surfels/hexels`) was not browsed directly this session beyond the "Hexel Edge" subdirectory described on the project page; worth checking its issue tracker and commit history in a future pass once it matures past "under development," both for third-party engagement and for whether the core training/benchmark code is ever released.
- No Hacker News, Reddit, or X/Twitter discussion of Hexels was found as of 2026-10-29 — worth a repeat search once the paper has had more time to circulate.

## Entity candidates

- Binbin Huang — person — lead author of [[claim-hexels-replace-gaussian-kernel-with-hexagonal-support-and-linear-color-field-in-2dgs|2D Gaussian Splatting]] (SIGGRAPH 2024/ACM TOG), the older, peer-reviewed baseline Hexels measures its headline PSNR gains against — flagged first, per the blind-spot note, since this is the foundational prior work the Hexels comparison actually rests on.
- Jan Held — person — lead author of Triangle Splatting (3DV 2026, peer-reviewed), the second baseline Hexels is benchmarked against.
- 2D Gaussian Splatting (2DGS) — concept — the primary baseline method across all three promoted Hexels claim-notes; already wikilinked but not yet its own concept note.
- Triangle Splatting — concept — the second baseline method; newly confirmed peer-reviewed (3DV 2026) this session.
- Prakhar Kaushik — person — Hexels first author, CCVL lab, Johns Hopkins.
- Soumava Paul — person — Hexels co-author.
- Alan Yuille — person — Hexels senior author / PI, Johns Hopkins CCVL lab.
- Thomas C. Hales — person — proved the honeycomb conjecture (2001), cited in the Hexels paper as the mathematical justification for the hexagon-over-ellipse design choice; already mentioned in a prior claim-note's body but not yet an entity page.