What consistency and quality evaluations have the authors of HunyuanWorld 1.0, WonderWorld, WonderJourney and World Labs published, and can their numbers be compared?

Each source evaluates differently and none measures 3D geometric consistency directly. HunyuanWorld 1.0 reports no-reference image-quality metrics (BRISQUE, NIQE, Q-Align) and CLIP alignment on rendered views against baselines, with no ground truth; WonderWorld's only quantitative table is generation time (9.5 s vs 749.5 s and 798.1 s on an A6000) plus qualitative comparison and a stated failure mode; WonderJourney reports a human preference study (400 participants, binary choices on four axes). The World Labs pages read give examples and claims but no published metric; its TOS disclaims dimensional accuracy. The tables are not comparable across papers (different views, baselines, protocols). Curator correction 2026-10-24 (discovery audit record C3-B1039-01, curator-reviewed C3R-0686; excerpts re-located by independent curator GET 2026-10-24): Correction. Served record C3-B012-W4-04 says WonderWorld's only quantitative table is generation time (9.5 s vs 749.5 s and 798.1 s on an A6000) plus qualitative comparison, citing arXiv v1. The record's summary did not scope the claim to a version (only its evidence and systems fields name arXiv v1, submitted 13 Jun 2024), and v4 (revised 25 Mar 2025, before the record was checked on 2026-10-22) was already current, so the claim holds for v1 but not for v4. arXiv lists v4 revised 25 Mar 2025 with the comment "CVPR 2025", and v4 adds Table 2 (novel-view renderings: CLIP score, CLIP consistency, CLIP-IQA+, Q-Align, CLIP aesthetic) and Table 3 (human 2AFC preference on bird-eye view rendering against WonderJourney, LucidDreamer and Text2Room). These are CLIP-based and preference metrics; the paper presents CLIP consistency as novel-view consistency, and, by inference (the paper does not say so), not a direct 3D geometric metric, so the record's claim that no source measures 3D geometric consistency directly is not contradicted. Generation-time numbers are unchanged in v4 (749.5 s, 798.1 s, 766.9 s, 9.5 s). Anyone comparing WonderWorld with the other evaluations should cite v4 and not describe WonderWorld as having no quantitative quality metrics. Curator qualifications: Completes C3R-0211 (README vs paper Director3D 3.2 vs 2.7 conflict kept). All strands verified. No published metric on the World Labs blogs read: a case-sensitive 'FID' search gives 0 hits, and the case-insensitive hits were 'fidelity'. HunyuanWorld's WonderJourney baseline uses HunyuanWorld's own right-90-degree trajectory, so it is not WonderJourney's protocol. Numbers are not comparable across papers. Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.

{
  "citations": [
    {
      "url": "https://arxiv.org/abs/2507.21809v2",
      "note": "Table 2; limitations; Sec 4 human study; Sec 3.2 protocol; OUTPUT LIMITATIONS"
    },
    {
      "url": "https://github.com/Tencent-Hunyuan/HunyuanWorld-1.0/blob/57fa9f3a79eae1079c279968cdb9d82c75fa7c86/README.md",
      "note": "Table 2; limitations; Sec 4 human study; Sec 3.2 protocol; OUTPUT LIMITATIONS"
    },
    {
      "url": "https://arxiv.org/abs/2406.09394v1",
      "note": "Table 2; limitations; Sec 4 human study; Sec 3.2 protocol; OUTPUT LIMITATIONS"
    },
    {
      "url": "https://openaccess.thecvf.com/content/CVPR2024/papers/Yu_WonderJourney_Going_from_Anywhere_to_Everywhere_CVPR_2024_paper.pdf",
      "note": "Table 2; limitations; Sec 4 human study; Sec 3.2 protocol; OUTPUT LIMITATIONS"
    },
    {
      "url": "https://www.worldlabs.ai/terms-of-service",
      "note": "Table 2; limitations; Sec 4 human study; Sec 3.2 protocol; OUTPUT LIMITATIONS"
    },
    {
      "url": "https://www.worldlabs.ai/blog/marble-world-model",
      "note": "Table 2; limitations; Sec 4 human study; Sec 3.2 protocol; OUTPUT LIMITATIONS"
    }
  ]
}