Abstract:We introduce PEPO (Pessimistic Ensemble based Preference Optimization), a single-step Direct Preference Optimization (DPO)-like algorithm to mitigate the well-known over-optimization issue in preference learning without requiring the knowledge of the data-generating distribution or learning an explicit reward model. PEPO achieves pessimism via an ensemble of preference-optimized policies trained on disjoint data subsets and then aggregates them through a worst case construction that favors the agreement across models. In the tabular setting, PEPO achieves sample complexity guarantees depending only on a single-policy concentrability coefficient, thus avoiding the all-policy concentrability which affects the guarantees of algorithms prone to over-optimization, such as DPO. The theoretical findings are corroborated by a convincing practical performance, while retaining the simplicity and the practicality of DPO-style training.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2602.06239 [cs.LG] |
| (or arXiv:2602.06239v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2602.06239 arXiv-issued DOI via DataCite |
Submission history
From: Luca Viano [view email]
[v1]
Thu, 5 Feb 2026 22:31:07 UTC (907 KB)
[v2]
Wed, 13 May 2026 16:34:43 UTC (916 KB)