PHILIA-EcoSensory Swarm v48 main: PPG as Temporal Stabilizer — 40-Seed Independent Replication and F4 Canonicalization

We present v48 main, an independent replication study validating the functional role of the Pulse Generator (PPG) as a temporal stabilizer of coherence in PHILIA-EcoSensory Swarm. Using an independent seed set (62–101, N = 40 per level, 240 runs total, T = 80,000 steps each), we demonstrate that coherence (F6) reliably emerges in L3 but remains transient in the absence of PPG (PASS★ = 0% in both datasets). The introduction of PPG (L3→L4) transforms this transient coherence into a persistent state, increasing strict PASS★ rates from 0% to 77.5% (synthetic) and 0% to 82.5% (real), with large effect sizes (Cohen's d = 3.26–3.70, p < 0.0001). Additionally, we resolve the previously reported 16× F4 discrepancy by introducing a dual formulation: F4a (population mean diversity, tail 5,000-step mean = 0.3336) reproduces v47.4 div = 0.331 within 95% CI, while F4b (temporal variation, mean = 0.020) is remarkably stable across all tested levels within experimental variance. The F4a/F4b ratio (~16× synthetic, ~19× real) is structurally invariant across module levels. Real data L5 achieves F6 = 0.891 ± 0.015, reproducing v48a (0.895 ± 0.016) to within 0.4%. These results establish a principled distinction between coherence existence and coherence persistence, positioning PPG as a minimal stabilizing module within the PHILIA architecture. All numerical results from local machine execution (AMD Ryzen 9 9800X3D + RTX 4080 SUPER 16GB). No proxy values used. 실측이 생명. 증명이 아닌 기술. — Trinity AI Research Team | April 2026