Construct randomized summaries for conformal configuration screening

Construct and analyze randomized summary states for conformal screening of post-hoc quantile-averaged ensemble configurations, including their validity, approximation quality, mergeability, and size.

Background

The lower-bound analysis in the paper concerns deterministic states that dominate the exact conformal shift for every multiset. Randomized summaries are outside the state classes analyzed by those deterministic lower bounds.

The conclusion explicitly lists randomized summaries as an open direction, leaving unresolved whether randomization can provide smaller or more accurate mergeable summaries while retaining the required conformal guarantees.

References

Open are adaptive recalibration under drift, the exponent gap between the general and net-histogram lower bounds, and randomized summaries.

— CALQA: Conformal Screening of Forecast-Ensemble Configurations from One Mergeable State  (2610.13249 - Liu et al., 9 Oct 2026) in Section 6, Conclusion