The Offline-Frontier Shift: Diagnosing Distributional Limits in Generative Multi-Objective Optimization

TL;DR

This paper investigates the limitations of generative models in offline multi-objective optimization, identifying distributional shifts as a key challenge and proposing diagnostic methods to understand their failure modes.

Contribution

It introduces the concept of offline-frontier shift as a fundamental limitation and provides empirical diagnostics for generative methods' performance in offline MOO.

Findings

Generative methods underperform evolutionary algorithms on metrics like generational distance.

Offline-frontier shift causes displacement of datasets from the Pareto front, limiting optimization.

Empirical evidence shows generative models tend to stay close to offline objective distributions.

Abstract

Offline multi-objective optimization (MOO) aims to recover Pareto-optimal designs given a finite, static dataset. Recent generative approaches, including diffusion models, show strong performance under hypervolume, yet their behavior under other established MOO metrics is less understood. We show that generative methods systematically underperform evolutionary alternatives with respect to other metrics, such as generational distance. We relate this failure mode to the offline-frontier shift, i.e., the displacement of the offline dataset from the Pareto front, which acts as a fundamental limitation in offline MOO. We argue that overcoming this limitation requires out-of-distribution sampling in objective space (via an integral probability metric) and empirically observe that generative methods remain conservatively close to the offline objective distribution. Our results position offline MOO as a distribution-shift--limited problem and provide a diagnostic lens for understanding when and why generative optimization methods fail.