33.65 turns the beta-minus/beta-plus pair into a tightening court: with the same weak-probe chain, the same axis, frozen clustered same-window selection above Pth, and blinded channel/axis labels, Corr(p_miss∥, Δt_common) and the fitted slope a must flip sign between channels, satisfy a(beta-minus) ≈ −a(beta-plus) within uncertainty, remain self-consistent under axis reversal, disappear in pulse-off / below-threshold / permutation controls, and stay nondispersive across bands.
Use this section as a compact machine-readable EFT reference.
33.65 turns beta-minus and beta-plus into one mirror-sign arbitration court. The admissible claim is not that either channel alone proves anything about neutrinos or antineutrinos, but that under one fixed chain and one fixed axis the nondispersive common-term step forms an opposite-sign mirror across the paired channels.
Measurement preserves one cross-channel mirror ledger. The hard outputs are Δt_common inside the event window under one frozen sign convention, missing momentum and its axial projection p_miss∥ under one preregistered closure aperture, Corr(p_miss∥, Δt_common), the fitted slope a and intercept b for each channel, and the mirror check that a(beta-minus) plus a(beta-plus) remains approximately zero within uncertainty. Only clustered events that already pass the 33.63/33.64 same-window fingerprint and lie at or above Pth may enter the court.
The workflow freezes the paired comparison before any verdict. Beta-minus and beta-plus data are acquired under the same weak-probe chain, the same clock reference, the same axis definition, the same symmetric detector geometry, and the same analysis pipeline; Pth, the event-window duration ΔT, clustering rules, and exclusion rules are shared across both channels; channel labels, axis-direction labels, and batch identifiers stay blinded during correlation and slope extraction; and an interleaved positive-axis / negative-axis / zero sequence suppresses slow drift under one fixed settling time.
Controls must break false mirror structure rather than merely weaken it cosmetically. With P = 0 or P < Pth, the opposite-sign correlation and slope pattern may not stabilize; channel-label or axis-label permutations must collapse the mirror relation; and Δt_common must remain nondispersive across at least two bandwidths or frequency bands, otherwise the effect returns to medium or instrument ledgers.
Support requires three things at once: clustered same-window events above Pth must show stably opposite correlation signs between channels, the fitted slopes must satisfy a(beta-minus) approximately equals minus a(beta-plus) within uncertainty, and axis reversal must flip both slopes in tandem while preserving the mirror relation. Falsification follows from no stable opposite sign, collapse under axis reversal, null tests that reproduce the same-strength mirror, or loss of nondispersion. The named adversaries are weak-probe crosstalk, trigger-synchronized pseudo-correlation, axis-definition drift, and closure-aperture calibration bias.
So the chapter closes only one tightened beta-minus/beta-plus common-term mirror-sign ledger aligned with the V03/V08/V09 audit lane. “Neutrino” and “antineutrino” mirror-step language may survive only as cross-channel arbitration names for the opposite-sign slope structure, not as carrier ontology, transport ontology, or canon-core verdicts. Its clean onward value is to hand a same-chain transport test to 33.66 and a spatial cone-profile audit to 33.67.