2026-08-15 · 3D candidate; pole gate not accepted
Paired full-vector 3D transfer
qualified artifacts updated 2026-08-15 11:11 UTC
Initial and optimized 2.5D densities were fitted independently to the same 3D topology map and their vector modes were reacquired on a common Yee grid.
Vector overlap
0.997021
Peak gap response
+4.453%
Driven response
+5.029%
Apparent pole Q
+1.440%
Initial pole gate
FAIL
Optimized pole gate
FAIL
Decision. Mode identity and the direct centered-gap response improvement are credible. The apparent 1.44% pole-Q gain is not accepted because the two-mode cluster fails strict wide/local and holdout gates.
Gate record
| Gate | Result |
|---|---|
| holdout q range ≤ 0.15 | FAIL |
| linewidth q difference ≤ 0.10 | PASS |
| maximum holdout pole shift ≤ 0.15 halfwidth | FAIL |
| maximum holdout response nrmse ≤ 1e-5 | PASS |
| simo loewner q difference ≤ 0.10 | PASS |
| wide local pole shift ≤ 0.15 halfwidth | FAIL |
| wide local q difference ≤ 0.05 | PASS |
Reproducibility artifacts
Reports, raw metrics, configurations, checkpoints, and validation data retained for this attempt.