Guardado en:
Detalles Bibliográficos
Autores principales: Bajgar, Ondrej, Gould, Dewi S. W., Liu, Jonathon, Abate, Alessandro, Gatsis, Konstantinos, Osborne, Michael A.
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2508.03693
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866918144010878976
author Bajgar, Ondrej
Gould, Dewi S. W.
Liu, Jonathon
Abate, Alessandro
Gatsis, Konstantinos
Osborne, Michael A.
author_facet Bajgar, Ondrej
Gould, Dewi S. W.
Liu, Jonathon
Abate, Alessandro
Gatsis, Konstantinos
Osborne, Michael A.
contents As AI systems become increasingly autonomous, reliably aligning their decision-making with human preferences is essential. Inverse reinforcement learning (IRL) offers a promising approach to infer preferences from demonstrations. These preferences can then be used to produce an apprentice policy that performs well on the demonstrated task. However, in domains like autonomous driving or robotics, where errors can have serious consequences, we need not just good average performance but reliable policies with formal guarantees -- yet obtaining sufficient human demonstrations for reliability guarantees can be costly. Active IRL addresses this challenge by strategically selecting the most informative scenarios for human demonstration. We introduce PAC-EIG, an information-theoretic acquisition function that directly targets probably-approximately-correct (PAC) guarantees for the learned policy -- providing the first such theoretical guarantee for active IRL with noisy expert demonstrations. Our method maximises information gain about the regret of the apprentice policy, efficiently identifying states requiring further demonstration. We also present Reward-EIG as an alternative when learning the reward itself is the primary objective. Focusing on finite state-action spaces, we prove convergence bounds, illustrate failure modes of prior heuristic methods, and demonstrate our method's advantages experimentally.
format Preprint
id arxiv_https___arxiv_org_abs_2508_03693
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PAC Apprenticeship Learning with Bayesian Active Inverse Reinforcement Learning
Bajgar, Ondrej
Gould, Dewi S. W.
Liu, Jonathon
Abate, Alessandro
Gatsis, Konstantinos
Osborne, Michael A.
Machine Learning
As AI systems become increasingly autonomous, reliably aligning their decision-making with human preferences is essential. Inverse reinforcement learning (IRL) offers a promising approach to infer preferences from demonstrations. These preferences can then be used to produce an apprentice policy that performs well on the demonstrated task. However, in domains like autonomous driving or robotics, where errors can have serious consequences, we need not just good average performance but reliable policies with formal guarantees -- yet obtaining sufficient human demonstrations for reliability guarantees can be costly. Active IRL addresses this challenge by strategically selecting the most informative scenarios for human demonstration. We introduce PAC-EIG, an information-theoretic acquisition function that directly targets probably-approximately-correct (PAC) guarantees for the learned policy -- providing the first such theoretical guarantee for active IRL with noisy expert demonstrations. Our method maximises information gain about the regret of the apprentice policy, efficiently identifying states requiring further demonstration. We also present Reward-EIG as an alternative when learning the reward itself is the primary objective. Focusing on finite state-action spaces, we prove convergence bounds, illustrate failure modes of prior heuristic methods, and demonstrate our method's advantages experimentally.
title PAC Apprenticeship Learning with Bayesian Active Inverse Reinforcement Learning
topic Machine Learning
url https://arxiv.org/abs/2508.03693