Run · Voynich Manuscript

T42 generation 1: neural posterior estimation on the T39 draws, calibrated against rejection ABC

Neural posterior estimation for the period-cipher family, with simulation-based calibration: SBC coverage of 90 % sets per setting; mean log posterior of the true setting (NPE vs ABC); posterior predictive misses on held-out pages.

Passes its gate (calibrated, beats ABC on the planted settings). The posterior predictive misses fewer held-out statistics than T39's best, but a posterior predictive spreads over many settings and so is easier to fall inside than one setting's 10 seeds; the like-for-like check (mode setting, 10 seeds) is generation 2. SBC rank test flags p_break (p 0.002) as miscalibrated. Remaining misses in both sections: glyph predictability (h2) and one-off words (hapax), as in T39.every setting covered on 96-100 % of draws; mean log score NPE -0.784 vs ABC -0.968. Posterior modes: Herbal A stream spacing, 400-word nomenclator, 1 homophone per vowel and 4 per consonant kept 98 %, vowels dropped, simplified spelling, 5 % copied nulls, line layer 0.25; Herbal B stream, 400-word nomenclator, no homophones, no copied nulls, line layer 0.5. Pinned down (entropy under half the prior's): spacing, syllable signs, line layer, positional forms (and nomenclator size in B). Posterior predictive on held-out pages: 2/12 held-out statistics missed (A), 1/12 (B)0

Result

Run	Measured	Result	Against	Decision	Date
Run	SBC coverage of 90 % sets per setting; mean log posterior of the true setting (NPE vs ABC); posterior predictive misses on held-out pages	every setting covered on 96-100 % of draws; mean log score NPE -0.784 vs ABC -0.968. Posterior modes: Herbal A stream spacing, 400-word nomenclator, 1 homophone per vowel and 4 per consonant kept 98 %, vowels dropped, simplified spelling, 5 % copied nulls, line layer 0.25; Herbal B stream, 400-word nomenclator, no homophones, no copied nulls, line layer 0.5. Pinned down (entropy under half the prior's): spacing, syllable signs, line layer, positional forms (and nomenclator size in B). Posterior predictive on held-out pages: 2/12 held-out statistics missed (A), 1/12 (B)	T39 rejection ABC best settings: 5/12 (A), 3/12 (B) on held-out pages with one setting and 10 seeds	Passes its gate (calibrated, beats ABC on the planted settings). The posterior predictive misses fewer held-out statistics than T39's best, but a posterior predictive spreads over many settings and so is easier to fall inside than one setting's 10 seeds; the like-for-like check (mode setting, 10 seeds) is generation 2. SBC rank test flags p_break (p 0.002) as miscalibrated. Remaining misses in both sections: glyph predictability (h2) and one-off words (hapax), as in T39.

What it used and found

Neural posterior estimation for the period-cipher family, with simulation-based calibration

Method

MTH

Herbal A and Herbal B paragraph text: training and held-out pages

Dataset

DAT

Latin genre texts

Dataset

DAT

ZL3b transcription

Dataset

DAT

Latin as a re-spaced homophonic stream (period practices)

Answer it tests

ANS

Parameters

abc = nearest 5 % of same-target draws, MAD scale, ppc = 20 settings drawn from the factorised posterior, seeds 7000-7019, held-out pages, draws = round-1 T39 draws only (3,400: 1,700 per herbal target), folds = 5, network = MLPClassifier(64, 64), alpha 1e-3, early stopping, one per setting

Cite it

Mysteries Department. (2026). T42 generation 1: neural posterior estimation on the T39 draws, calibrated against rejection ABC [Run RUN0DCF8MBKZFHJ]. https://mysteriesdept.com/r/RUN0DCF8MBKZFHJ

@misc{mysteriesdept_run0dcf8mbkzfhj,
  title     = {T42 generation 1: neural posterior estimation on the T39 draws, calibrated against rejection ABC},
  author    = {{Mysteries Department}},
  year      = {2026},
  note      = {Run, catalog key RUN0DCF8MBKZFHJ},
  url       = {https://mysteriesdept.com/r/RUN0DCF8MBKZFHJ}
}