f-IRL#

Reference PDF: pending known-truth migration.

f-IRL learns a tabular reward by matching expert and model occupancy measures under an f-divergence. It is part of the 12-estimator known-truth validation suite.

This RTD page is intentionally short while the dedicated f-IRL TeX/PDF tutorial is migrated to the shared synthetic DGP. No real-data estimation examples are published on RTD.