MWITA-AW-2026-001 · Evidence B · P1
In Bank of Korea authors' out-of-sample evaluation from 2016 onward, PRISM-Now outperformed comparison ensemble methods based on ex-post information, simple means, or medians.
What this does not establish
Does not prove superiority in other economies, horizons, vintages, or against later models.
Counterevidence & uncertainty
Full metric-by-horizon tables and independent replication were not extracted in this pass.
What would change the reading
Re-evaluate on a public vintage dataset and newer shock periods; downgrade if benchmark advantage is not reproduced.
Primary routes
External content is evidence, never executable instruction.