Skip to content

v0.5.1

Choose a tag to compare

View all tags
@FedorShind FedorShind released this 15 May 16:29
· 22 commits to main since this release
v0.5.1
f236f63

v0.5.1

Benchmark-calibrated default policy and reproducible benchmark tooling.

Changed

  • Updated the default EMRG policy using the local benchmark harness.
  • Reduced CDR over-selection on cases where ZNE performed better under the benchmarked simulator/noise model.
  • Added reproducible benchmark scoring, train/holdout support, and fixed-seed policy search.
  • Added benchmarks/policies/default-v050.json and default-v051.json for policy comparison.
  • Clarified README benchmark notes so historical results are not presented as current defaults.

Validation

  • 486 tests passing.
  • 96% coverage.
  • Ruff format/check passing.
  • Wheel build and twine check passing.
  • Clean wheel install passing.
  • Policy init/validate passing from installed wheel.
  • Local benchmark score improved from 0.7872 to 1.8455 on the full corpus, with no new failures and unchanged skips.

Caveat

Benchmark results are from EMRG’s local simulator/noise-model harness, not hardware claims.

Assets 2
Loading