Narrower radial smoothing is not a monotone improvement
The validated numerical baseline stays unchanged. New local controls show why choosing its radial smoothing policy needs a separate physics decision. The primer explains the guarded surrogate and fallback model.
We evaluated the same retained plasma states with four radial smoothing widths. Transition widths, guards, clipping, floors, patches and profiles stayed fixed. The native operator includes the transport coefficients after processing, not only the surrogate’s raw particle flux.
At the particle-split boundary, smaller nonzero radii do not move the result steadily toward the raw target: the selected particle rate is 0%, 6.76754%, 9.05578% and 5.76626% less inward than the raw target at radial HWHM 0, 0.025, 0.05 and 0.1 respectivelyC-312. The selected response to a density perturbation also changes: the target derivative is +39.9009; native values are +39.9009, +0.4076, −43.8378 and −38.8539 across the same widths; the +0.4076 value lacks strict finite-difference corroborationC-313. A small AD value without the required finite-difference agreement remains unresolved.
Zero radial width recovers the raw target at that particle state. It does not solve every mechanism. At the coefficient-cap context, the guarded served flux already differs before radial processing: the selected cap rate is already 96.87350% less inward than the raw target without radial smoothing, compared with 81.44208%, 63.19414% and 65.09036% at the three nonzero widthsC-314. That raw-target difference is not a measured physical error, and it must not be attributed entirely to smoothing.
The saved-state reconstructions retain the original baseline control and all derivative qualifications: 696 observations pass algebra/invariant reconstruction and 174 baseline observations reproduce the parent evidence; 996 of 1008 derivative components and 45 of 48 complete direction sets meet the strict ruleC-315. A failed cap process and a replay checker correction remain in the evidence history; the numerical equations and acceptance thresholds were not relaxed.
These are local operator comparisons, not five-second density or fusion-power errors. They support a constrained particle-flux comparison through the full postprocessing operator before any trajectory policy changes. Valid-trajectory matched windows and an explicit density/flux accuracy requirement are still needed to judge accumulated effects.
No production width was selected, no batch job was submitted and physicist sign-off remains pending.
Technical details → radial controls and limitations. Milestone → M5.
Technical details → Uncertainty propagation and validation › radial smoothing controls
Decisions → OQ-47