Run · Voynich Manuscript
T41 generation 6: 400 restarts on the one-sign-per-letter control; better likelihood, still under the bar
Lattice EM decipherment: sign boundaries and key learned together, with a structure-preserving null: Best restart's training log-likelihood; key accuracy; held-out dictionary coverage (bar 0.483).
Result
| Run | Measured | Result | Against | Decision | Date |
|---|---|---|---|---|---|
| Run | Best restart's training log-likelihood; key accuracy; held-out dictionary coverage (bar 0.483) | ll -40,852 (best restart 205, seed 6205), key 0.623, coverage 0.340, character similarity 0.254 | 200 restarts: ll -40,969, key 0.698, coverage 0.374 · EM from the true key: ll -40,505, coverage 0.518 | Fails the control gate. Doubling the restarts gains 117 nats of likelihood but not key accuracy or coverage; the gap to the true key's basin (350 nats) shrinks only logarithmically with restarts. Generation 7 searches around the best restart instead (basin hopping). |
What it used and found
MTH
Latin genre texts
DATDataset
Latin as a re-spaced homophonic stream (period practices)
ANSAnswer it tests
Parameters
pw = 0.25, alpha = 0.05, restarts = 400, settings = one sign per letter (as g5c), permutations = 10, em_iterations = 60, restart_seeds = 6000-6399
Cite it
Mysteries Department. (2026). T41 generation 6: 400 restarts on the one-sign-per-letter control; better likelihood, still under the bar [Run RUN0DCHDDGSSB3H]. https://mysteriesdept.com/r/RUN0DCHDDGSSB3H
@misc{mysteriesdept_run0dchddgssb3h,
title = {T41 generation 6: 400 restarts on the one-sign-per-letter control; better likelihood, still under the bar},
author = {{Mysteries Department}},
year = {2026},
note = {Run, catalog key RUN0DCHDDGSSB3H},
url = {https://mysteriesdept.com/r/RUN0DCHDDGSSB3H}
}