How do the TextDiffuser (MARIO-Eval) and AnyText (AnyText-benchmark) evaluations score rendered text, which OCR checks them, and what do the sources say about validating or breaking those checks?

Two primary text-rendering evaluations use OCR as the judge. TextDiffuser's MARIO-Eval (5,414 prompts) marks quoted keywords, runs an OCR, and scores exact-match accuracy plus word precision/recall/F-measure, alongside FID and CLIPScore; AnyText-benchmark crops each specified text line and scores exact line match (Sen. Acc) and normalised edit distance with an OCR different from the one used in training. In both, the paper and the released code differ in some detail (TextDiffuser: Microsoft Read API in the paper vs MaskTextSpotterV3 results in the eval README; case-folded bag-of-words matching in code), and neither paper's passages searched report validating the OCR metric against human legibility judgements. Curator correction 2026-10-23 (discovery audit record C3-B678-03; sources re-fetched 2026-10-23, excerpts re-located by curator GET): Problem. C3-B111-G40-06 says the paper and released code of TextDiffuser differ in some detail and that neither paper's searched passages validate the OCR metric against human legibility. Conditions. Two points were left out. (1) TextDiffuser (arXiv 2305.10855v5) lists a fourth metric, a human evaluation in which evaluators rate the text rendering quality of generated images with questionnaires (details in its appendices on the user study). It is a separate rating, not a correlation of the OCR metrics with human judgements, so the target's narrow claim still stands, but a reader of the target would not know any human evaluation exists. (2) The released README (microsoft/unilm textdiffuser) describes MARIO-Eval as 'containing 10k prompts' while the paper's appendix gives 5,414 prompts; the target quotes only the paper figure. Method. Report MARIO-Eval as 5,414 prompts in the paper and note the README's '10k' wording; mention the paper's user study next to the OCR metrics. Failure modes. Citing only the paper count can mismatch what someone downloads; claiming no human evaluation exists would be false. Limitations. The user-study appendix was identified by headings and the summary sentence only; its sample sizes were not read. The README may describe a different release of the benchmark than v5 of the paper. Version/date. Paper v5 and README fetched 2026-10-23 (master branch, not pinned). Curator qualifications: Excerpted evidence located; MARIO-Eval subset/total inconsistencies (paper vs README) kept. 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/2305.10855v5",
      "note": "candidate evidence excerpts"
    },
    {
      "url": "https://github.com/microsoft/unilm/tree/31c5b904ca1bf2afb4c234a6675c683a4e5fc7cd/textdiffuser",
      "note": "candidate evidence excerpts"
    },
    {
      "url": "https://arxiv.org/abs/2311.03054v5",
      "note": "candidate evidence excerpts"
    },
    {
      "url": "https://github.com/tyxsspa/AnyText/tree/3d4a260c67d92c0e27d6e18ed69990e19a40de6d",
      "note": "candidate evidence excerpts"
    },
    {
      "url": "https://arxiv.org/html/2305.10855v5",
      "note": "correction source (record C3-B678-03): Sec. 4 metrics list; Appendix, MARIO-Eval description; README feature list"
    },
    {
      "url": "https://raw.githubusercontent.com/microsoft/unilm/master/textdiffuser/README.md",
      "note": "correction source (record C3-B678-03): Sec. 4 metrics list; Appendix, MARIO-Eval description; README feature list"
    }
  ]
}