Does the new energy model reproduce published ordering temperatures for nine alloys?

E190 — rung 1 from v5, scored against the published ordering table

forager/ladder/ece_odt.py::transition_ece asks v5 for an ordering temperature by pyeCE canonical Monte Carlo (annealed 1800 → 300 K, 12 states, 4×4×4 = 128 sites, T_c from the heat-capacity peak with a parabolic refinement; sigma = half-width at 70%, floored at the sweep step) and returns what rungs.transition returns. Validation before it replaces anything: every distinct composition in published_odt.PUBLISHED_ODT (Sobieraj 2020, Fernández-Caballero 2017, Woodgate 2023, Kim & Widom 2023 — all DFT-CE + MC, the same quantity by the same route), one checkpoint each, load-gated. Predictions. (Two entries — WAK23 MoNbTa, KW23 MoNbTaW — report no transition; they are scored separately: v5 should find none above 300 K or a weak one, and either is stated.)

1. ≥ 7 of the 10 numeric entries within ±200 K of the published value (the table's own method spread is 100 K; a 128-site cell and 12 states add ~±75 K of resolution). 2. Falsified at ≤ 4 of 10: v5's ordering energies do not transfer across the table and rung 1 stays on hold — then the failure pattern (which chemistries) is the finding. 3. The Mo–Ta-containing entries (KW23, FC17) land below the icet CE's numbers by ≥ 2× — E188's result generalises.
2. Wall time ≤ 10 min per composition at 4 threads.

E190 interim (15:29) — seven of seven low, chemistry-independent. At the validation resolution (128 sites, 12 states, 60 samples): CrTaTiVW ≤ 300 (pub. 900/1000), TaTiVW ≤ 300 (500), CrTaTiW ≤ 300 (500), CrTaVW 435 (1200/1300), CrTiVW 425 (700), CrTaTiV ≤ 300 (600), MoNbTaVW ≤ 300 (750, 742). The last is the decisive one: a refractory-centred alloy whose formation energies v5 gets within 10 meV/atom, and it still reads "Cv still rising at 300 K". Two readings, one test between them: (i) v5's ordering energies — the differences between decorations at fixed composition, a much smaller quantity than the formation energy — are weak across the board; (ii) the validation sweep is too coarse to grow a heat-capacity peak (E188's clean Mo–Ta peak took 432 sites and 100 samples on the strongest-ordering pair in the table).

E190 result (15:31) — prediction 1 FALSIFIED, prediction 2 fires. All nine compositions at the validation resolution (runs/e190_odt_validation/table.json):

alloy               published (K)      v5 (K)   read
CrTaTiVW            900 / 1000         ≤ 300    still rising at the coldest state
TaTiVW              500                ≤ 300    censored
CrTaTiW             500                ≤ 300    censored
CrTaVW              1200 / 1300        435      peak; 800 low
CrTiVW              700                425      peak; 275 low
CrTaTiV             600                ≤ 300    censored
MoNbTaVW            750 / 742          ≤ 300    censored — the DFT-validated chemistry
MoNbTa              none reported      583      a peak where WAK23 reports none
MoNbTaW             none reported      ≤ 300    consistent with "none"

The script's count "2 of 10 within 200 K" is the two censored ≤ 300 reads against published 500 K — not hits. Scored honestly: 0 of 10. Every v5 temperature is low, by 275 K to ≥ 900 K, in every chemistry including the one whose formation energies v5 reproduces to 10 meV/atom. So rung 1 does not transfer from v5 as fitted and as swept here. Which of the two it is — v5's within-composition ordering energies being weak (RHEA's ordered cells are 2–16 atoms; the fit's ρ 0.80 on them is a ranking, not a scale), or the 128-site / 60-sample sweep failing to grow a peak that a 432-site / 100-sample sweep grows (E188) — is what E190b decides, and E188's Mo–Ta 500 K stands only if E190b shows MoNbTaVW's peak at the finer resolution. Prediction 3 (Mo–Ta entries ≥ 2× below icet) is moot: v5 is below everything. Prediction 4 (≤ 10 min/composition) held (~7 min each). Until E190b: rung 1 stays on the icet CE with its known 3.5× excess on Mo–Ta, and no ladder verdict is re-scored on v5's T_c — including E188's flip of the Mo–Ta basin, which is now "provisional, pending E190b" rather than a finding.