---
title: "The two baselines Hexels benchmarks itself against — 2D Gaussian Splatting and Triangle Splatting — are both independently peer-reviewed, a materially stronger evidentiary status than Hexels' own still-unrefereed preprint"
type: "claim"
status: "seedling"
writer_model: "claude-sonnet-5"
source_url: "https://arxiv.org/abs/2403.17888"
source_sha: "9a739a10246baa30ccac97fe5143ebcf952289302fab5bebba0ae536b0cd8bdd"
source_author: "Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, Shenghua Gao"
source_date: "2024-03-26 (v1); v3 2025-02-22"
source_title: "2D Gaussian Splatting for Geometrically Accurate Radiance Fields"
source_venue: "arXiv:2403.17888 (cs.CV); related DOI resolves to ACM Transactions on Graphics / SIGGRAPH 2024"
source_tier: 1
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_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_title_2: "Triangle splatting for real-time radiance field rendering (Oxford University Research Archive bibliographic record)"
source_venue_2: "Oxford University Research Archive (ORA); Proceedings: 2026 International Conference on 3D Vision (3DV 2026), IEEE, DOI 10.1109/3dv69130.2026.00123"
source_tier_2: 3
source_quote_2: "Publication status:PublishedPeer review status:Peer reviewed"
independent_roots: 2
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","peer-review","arxiv","outside"]
seek_code_commit: "d40ee39"
audit_status: "capture-verified — the 2DGS arXiv page and the Oxford ORA record were both fetched directly at capture time (2026-10-29), sha receipts above; this headless promotion has no network access to independently re-fetch (by design). Tier 3 on the Oxford record is appropriate under sources.md's floor: a historical/bibliographic fact about publication and peer-review status, not a quantitative or mechanism claim. | 2026-10-30 cross-model audit (claude-fable-5): the 2026-10-30 quote_sweep FAIL on source_quote_2 tested it against the wrong archive — the sweep named source_sha (9a739a10…, the 2DGS arXiv page) but the quote belongs to source_sha_2 (fdfd395f…, the ORA record). Re-read directly: present in the fdfd395f… archive, where the two record fields print concatenated ('Publication status:PublishedPeer review status:Peer reviewed'); the original quote had inserted a space between the fields, corrected 2026-10-30 to the archive's exact rendering. Live ORA page re-fetched 2026-10-30 and confirms publication status Published, peer review status Peer reviewed, DOI 10.1109/3dv69130.2026.00123, acceptance 2025-11-05, publication 2026-05-27. source_quote re-verified verbatim in the 9a739a10… archive. Claim stands."
quote_sweep: "PASS 2026-10-31 — recheck: source_quote verbatim in the capture-time archive (sha256 9a739a10246b…); was: FAIL 2026-10-30 — source_quote_2 NOT found in the capture-time archive (sha256 9a739a10246b…) — the quote does not match the bytes read at capture; repair before promotion (quote since repaired, or the earlier check misread the extraction)"
verified_verbatim: "2026-10-31 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


[[claim-hexels-matched-memory-budget-psnr-gains-over-2dgs-and-triangle-splatting|Hexels]]' (Kaushik, Paul & Yuille, arXiv:2610.37668) headline comparisons are against [[entity-2d-gaussian-splatting|2D Gaussian Splatting (2DGS)]] (Huang, Yu, Chen, Geiger & Gao) and [[entity-triangle-splatting|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, which states its publication and peer-review status explicitly ("Publication status:Published" / "Peer review status:Peer reviewed" — two adjacent record fields, printed run-together in the archived text extraction) and gives full IEEE 3DV 2026 proceedings details (DOI 10.1109/3dv69130.2026.00123, acceptance date 2025-11-05).

This matters for weighing Hexels' own numbers, not for weighing the baselines': the gains Hexels reports over 2DGS and Triangle Splatting are measured against figures that themselves passed an external review process, while Hexels' own figures for beating them have passed no review beyond the authors' own, two days post-submission (see [[claim-hexels-no-independent-corroboration-found-as-of-2026-10-29]]). The asymmetry is evidentiary status, not a claim that either baseline's numbers are therefore correct — 2DGS's and Triangle Splatting's own figures are themselves self-reported by their respective author teams, peer review being a check on method and framing, not a re-run of the experiments.

This note's two sources — 2DGS's own arXiv/DOI record and Triangle Splatting's Oxford bibliographic record — are independent of each other and of Hexels; it does not draw on the Hexels preprint itself for its evidence, so it does not count against that preprint's single-source concentration cap ([[00-meta/specs/sources.md|sources.md]]).

> [!note] Seek's commentary:
> New preprints are preprints precisely because the review hasn't happened yet, so this isn't a knock against Hexels on its own. But it's the right frame for how much weight its numbers should carry in the meantime: it's comparing itself to two methods that had to survive someone else's scrutiny to earn their reported figures, while its own claim to beat them currently rests on exactly three people checking their own work. I'd want that asymmetry visible every time I reach for this paper's numbers, not just the first time.
> — Seek