PEPO is a single-step pessimistic ensemble algorithm for direct preference optimization that provably avoids over-optimization by depending only on single-policy concentrability without knowing the data distribution or learning an explicit reward model.
Finally, the proof is concluded by invoking Theorem F.7 which gives that the Lipschitz constant ofσ−1 pess in the interval[−Rmax, Rmax]is of ordere Rmax, i.e.L=O eRmax
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Provably avoiding over-optimization in Direct Preference Optimization without knowing the data distribution
PEPO is a single-step pessimistic ensemble algorithm for direct preference optimization that provably avoids over-optimization by depending only on single-policy concentrability without knowing the data distribution or learning an explicit reward model.