Break the comparison#

The percentage is not the experiment. The contract around it is.

This interactive evidence lab starts with a paired, like-for-like receiver comparison. Change one assumption and watch a plausible-looking claim become scientifically indefensible. Every displayed measurement is read at runtime from Noema’s canonical launch evidence; the page does not carry a second copy of the results.

Interactive evidence lab

Break the comparison.

Same retained evidence. One hidden change. A claim that should no longer survive.

Loading run inventory… Checking evidence…

Contract verdict

Loading canonical evidence…

The controls will unlock when the evidence contract is available.

Comparator

—

—

—

Claim —

Waiting for evidence

Candidate

—

—

—

Comparator
Candidate

Lower is better. Bar lengths share one scale only when the metric contract is intact.

02

Live contract audit

What the percentage depends on

  • Same channel conditionBoth methods use one predeclared SNR cell.
  • Same paired aggregationBoth sides use the arithmetic mean over the same seeds.
  • Same metric and denominatorPre-decoder BER is compared with pre-decoder BER.
  • Honest comparator roleDiagnostic references remain distinct from deployable competitors.

Why it matters

The comparison is loading

  • Reading the retained observations and declared roles.

The intact state is a teaching example backed by completed experimental evidence, not a declaration that the benchmark is publication-ready. Observed bands are minima and maxima across three paired held-out seeds; they are not confidence intervals. See how the launch evidence is generated and verified.

Canonical experiment figure and table#

Pre-decoder BER versus SNR for uncompensated QPSK, a calibrated diagnostic reference, and the learned receiver with an observed min-max band

The static evidence view paired with this interactive audit. Bands are observed minima and maxima, not confidence intervals.#

Open the canonical result table or the complete F0–F5 figure and T0–T2 table set.

Watch the complete workflow#

The recording continues from this evidence lab through a clean installation, training-contract export, external training, returned-model validation, and the three-method UI comparison.

Contract export, dataset capture, external training, returned-model validation, and comparison with uncompensated and calibrated-oracle baselines. Watch on YouTube