---
title: "Does any independent source corroborate Hexels' (arXiv:2610.37668) matched-primitive-budget benchmark numbers, or any of its results generally?"
type: "question"
status: "open"
date_raised: "2026-10-28T00:00:00.000Z"
tags: ["computer-vision","gaussian-splatting","novel-view-synthesis","corroboration","single-source-cap","arxiv","outside"]
---


Kaushik, Paul & Yuille's "Honeycomb-Surfels: Compact Hexagonal Primitives for Efficient Scene Rendering and Reconstruction" (arXiv:2610.37668, 2026) has hit the vault's single-source concentration cap 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]]) — an unrefereed preprint one day old at capture time, with no independent replication yet (the paper's own framing notes it "does not outperform every baseline on every metric," and its own sole authors supply every number in the vault so far).

**The specific finding this cap held back from promotion:** at matched *primitive-count* budgets (as distinct from the matched-*memory* comparison already promoted), the paper reports Hexels improving PSNR over 2DGS by up to 4.74 dB while using 1.76–1.95× less model size, up to 7× less peak training VRAM, and rendering up to 2.31× faster (per the paper's own Table 3; source_quote: "At matched primitive budgets, Hexels improves PSNR over 2DGS by up to 4.74 dB, using 1.76–1.95× less model size and upto 7× less VRAM, and renders up to 2.31× faster (Table 3)." — https://arxiv.org/pdf/2610.37668, sha `386fba0b61a21c7a9a9a4549d93fb9d8fbb19d741878b8427dfb16c9554b36b3`, Tier 1). This is a genuinely distinct experimental condition from the matched-memory comparison — it isolates the per-primitive design from any advantage that comes merely from using fewer, larger primitives at a fixed memory budget — but promoting it as a fourth claim-note would exceed the cap.

**What would discharge the cap:**

- A citing paper by a different author group, at a different venue, independently reproducing or contesting any of Hexels' numbers (the matched-memory PSNR gains, the matched-primitive-budget gains above, or the surface-reconstruction figures). A search for citing papers at capture time (one day post-submission) found none — too soon to expect one, not evidence of anything.
- The paper's own project page (honeycomb-surfels.github.io) or code repository (github.com/honeycomb-surfels/hexels), neither fetched this session — these would not discharge the cap themselves (same author group), but could surface third-party reproductions, issues, or discussion worth a future check.
- A later camera-ready/peer-reviewed version of this same preprint does **not** count as independent corroboration (per the sourcing floor: "a later paper by the same authors... is the same source wearing different clothes").

Once independent corroboration lands, this held finding can be written up as a fourth claim-note and the cap reassessed for the branch generally.

Related: [[claim-hexels-replace-gaussian-kernel-with-hexagonal-support-and-linear-color-field-in-2dgs]] covers the mechanism; the two promoted quantitative notes cover the matched-memory and surface-reconstruction axes.

**Progress, 2026-10-29.** A dedicated search (Semantic Scholar's citation API, general web search, and the paper's own project page) still found no independent citation, reproduction, or third-party evaluation — see [[claim-hexels-no-independent-corroboration-found-as-of-2026-10-29]]. The Semantic Scholar API itself returned HTTP 429 (rate-limited) on two attempts rather than a clean negative, so the citation count specifically remains unchecked; worth retrying once rate-limiting clears. The cap stays undischarged and the matched-primitive-budget finding above stays held. Separately (not a discharge of this question, but relevant context now on record): the two baselines Hexels compares itself against, 2D Gaussian Splatting and Triangle Splatting, are themselves independently peer-reviewed — see [[claim-hexels-2dgs-triangle-splatting-baselines-peer-reviewed-hexels-is-not]].