Speaker
Description
We test whether a fixed positive-semidefinite channel metric improves a liveness-aware recursive trigger in an idealized single 16-channel multi-PMT module. This is a controlled extension of the same waveform synthesis, 200 ns deadtime model, and scalar first-order recursion used in companion CHEP 2026 contribution #562, not an independent detector simulation. Reconstructed energy and direction from 1,024 selected IceCube IceTracks-DR2 events drive synthetic channel waveforms; 256 event-driver rows train one metric and 768 are held out. A two-fold coincidence comparator, the equal-weight recursive score, and the metric-weighted score are independently calibrated on 100,000 shared noise-only windows to a nominal false-trigger probability of 10⁻³. On 200 realizations of 64 held-out events at P_dead = 0.5, efficiencies are 0.721 ± 0.058, 0.805 ± 0.050, and 0.652 ± 0.059, respectively. The metric-minus-plain recursive difference is −0.153 (paired 95% CI: −0.159 to −0.146). Thus, this particular trained geometric score does not improve the plain recursive method; the result limits, rather than validates, a proposed metric extension. No measured channel waveforms, detector-level performance, or propagated state uncertainty are claimed.