Saved in:
Bibliographic Details
Main Authors: Müller, Nicola J., Oster, Moritz, Valera, Isabel, Hoffmann, Jörg, Gros, Timo P.
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.17544
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911525592104960
author Müller, Nicola J.
Oster, Moritz
Valera, Isabel
Hoffmann, Jörg
Gros, Timo P.
author_facet Müller, Nicola J.
Oster, Moritz
Valera, Isabel
Hoffmann, Jörg
Gros, Timo P.
contents Learning per-domain generalizing policies is a key challenge in learning for planning. Standard approaches learn state-value functions represented as graph neural networks using supervised learning on optimal plans generated by a teacher planner. In this work, we advocate for learning Q-value functions instead. Such policies are drastically cheaper to evaluate for a given state, as they need to process only the current state rather than every successor. Surprisingly, vanilla supervised learning of Q-values performs poorly as it does not learn to distinguish between the actions taken and those not taken by the teacher. We address this by using regularization terms that enforce this distinction, resulting in Q-value policies that consistently outperform state-value policies across a range of 10 domains and are competitive with the planner LAMA-first.
format Preprint
id arxiv_https___arxiv_org_abs_2603_17544
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Per-Domain Generalizing Policies: On Learning Efficient and Robust Q-Value Functions (Extended Version with Technical Appendix)
Müller, Nicola J.
Oster, Moritz
Valera, Isabel
Hoffmann, Jörg
Gros, Timo P.
Artificial Intelligence
Machine Learning
Learning per-domain generalizing policies is a key challenge in learning for planning. Standard approaches learn state-value functions represented as graph neural networks using supervised learning on optimal plans generated by a teacher planner. In this work, we advocate for learning Q-value functions instead. Such policies are drastically cheaper to evaluate for a given state, as they need to process only the current state rather than every successor. Surprisingly, vanilla supervised learning of Q-values performs poorly as it does not learn to distinguish between the actions taken and those not taken by the teacher. We address this by using regularization terms that enforce this distinction, resulting in Q-value policies that consistently outperform state-value policies across a range of 10 domains and are competitive with the planner LAMA-first.
title Per-Domain Generalizing Policies: On Learning Efficient and Robust Q-Value Functions (Extended Version with Technical Appendix)
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2603.17544