Reference Field

The full population of operators against which any individual or cohort is compared. The benchmark population.

All operator submissions across all platforms and time windows.

Field-level distributions, global percentiles, field statistics.

The reference field provides the ground truth for relative positioning — an operator's rank is only meaningful relative to the field.

Verify field statistics are stable as the population grows (convergence check).

Field-level medians, quartiles, distributions for each metric.

If field statistics are unstable (do not converge as population grows), the reference field is not reliable.

3,304 models, 17 platforms, field-analysis dataset with medians and quartiles.

The field is biased toward operators who submit data — not a random sample of all AI users.

SigRank field-analysis schema, leaderboard methodology. Architecture: mos2es.com/architecture.