Evidence reclassified
GUD benchmark claims split by what the artifact actually measures
The tracked artifact numerically checks bigeometric power-law identities and stores values from the archived k(L) formula. It does not contain a controlled speed benchmark or independent physical-scale validation.
Supported
- D_BG[xⁿ] = eⁿ is reproduced numerically for selected powers.
- The result is a constancy diagnostic for the published operator.
- The artifact can be rerun as a local algebraic check.
Not supported by this artifact
- 40,000× physical accuracy or universal speedups
- A calculus recommendation determined by length scale
- Proof that NNC preserves arbitrary physical invariants
Tracked evidence
simulations/gud_composite_benchmarks.py writes results/gud_composite_benchmarks.json. The k values in that file are formula evaluations, not optimizer discoveries or experimental observations.