Split the Differences, Pool the Rest: Provably Efficient Multi-Objective Imitation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Sheebaelhamd, Ziyad, Viano, Luca, Cevher, Volkan, Vernade, Claire
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914577261789184
author Sheebaelhamd, Ziyad
Viano, Luca
Cevher, Volkan
Vernade, Claire
author_facet Sheebaelhamd, Ziyad
Viano, Luca
Cevher, Volkan
Vernade, Claire
contents This work investigates multi-objective imitation learning: the problem of recovering policies that lie on the Pareto front given demonstrations from multiple Pareto-optimal experts in a Multi-Objective Markov Decision Process (MOMDP). Standard imitation approaches are ill-equipped for this regime, as naively aggregating conflicting expert trajectories can result in dominated policies. To address this, we introduce Multi-Output Augmented Behavioral Cloning (MA-BC), an algorithm that systematically partitions divergent expert data while pooling state-action pairs where no behavior conflict is observed. Theoretically, we prove that MA-BC converges to Pareto-optimal policies at a faster statistical rate than any learner that considers each expert dataset independently. Furthermore, we establish a novel lower bound for multi-objective imitation learning, demonstrating that MA-BC is minimax optimal. Finally, we empirically validate our algorithm across diverse discrete environments and, guided by our theoretical insights, extend and evaluate MA-BC on a continuous Linear Quadratic Regulator (LQR) control task.
format Preprint
id arxiv_https___arxiv_org_abs_2605_12000
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Split the Differences, Pool the Rest: Provably Efficient Multi-Objective Imitation
Sheebaelhamd, Ziyad
Viano, Luca
Cevher, Volkan
Vernade, Claire
Machine Learning
This work investigates multi-objective imitation learning: the problem of recovering policies that lie on the Pareto front given demonstrations from multiple Pareto-optimal experts in a Multi-Objective Markov Decision Process (MOMDP). Standard imitation approaches are ill-equipped for this regime, as naively aggregating conflicting expert trajectories can result in dominated policies. To address this, we introduce Multi-Output Augmented Behavioral Cloning (MA-BC), an algorithm that systematically partitions divergent expert data while pooling state-action pairs where no behavior conflict is observed. Theoretically, we prove that MA-BC converges to Pareto-optimal policies at a faster statistical rate than any learner that considers each expert dataset independently. Furthermore, we establish a novel lower bound for multi-objective imitation learning, demonstrating that MA-BC is minimax optimal. Finally, we empirically validate our algorithm across diverse discrete environments and, guided by our theoretical insights, extend and evaluate MA-BC on a continuous Linear Quadratic Regulator (LQR) control task.
title Split the Differences, Pool the Rest: Provably Efficient Multi-Objective Imitation
topic Machine Learning
url https://arxiv.org/abs/2605.12000