EML-A5EML_SYMBOLIC_REGRESSION_TEMPLATE_SEARCH_PASSfixed-template searchfull PySR: false

EML Symbolic-Regression Template Search

A candidate-only scan over fixed EML, standard, control, and baseline templates for the same Chebyshev psi residual fixture.

Templates
6
Samples
800
Gamma scan
5-40
Known gamma1
14.134725
review finding

The EML critical-line template remains a compact candidate and leads the simple complexity-adjusted score on this fixture, but the template search is intentionally ambiguous: a wrong-exponent control localizes slightly closer, and low-MSE baselines show that MSE alone is not evidence of structure.

promising_but_not_decisivecandidate score leader: eml_critical_one_nodelocalization leader: eml_wrong_exponent_03MSE leader: constant_baseline
EML gamma error
0.047682
Standard gamma error
0.205260
Wrong-exponent gamma error
0.030173
templates
TemplateFamilyNodesParamsGammaMSEGamma error
eml_critical_one_nodeeml1114.0870445.3392890.047682
standard_profiled_sqrt_cos_sinstandard5313.9294655.2530050.205260
eml_wrong_exponent_03negative_control1114.1045525.8144250.030173
eml_wrong_exponent_07negative_control1120.9504755.7694756.815750
plain_profiled_cos_sinnegative_control3314.0520265.4615350.082699
constant_baselinebaseline01n/a3.555752n/a
non-claims
  • This search does not prove RH.
  • This search does not discover zeta zeros.
  • This search is not a full PySR run.
  • This search does not prove an EML grammar theorem.
  • This search does not promote any Atlas entry publicly.
  • This search does not change Forge/compiler behavior.