Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Xu, Tengye, Sun, Yangting, Shen, Ziju, Chen, Guanqi, Fu, Zhen, yizhou, Chen, Chen, Hua, Pan, Jia
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2605.22123
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911704280989696
author Xu, Tengye
Sun, Yangting
Shen, Ziju
Chen, Guanqi
Fu, Zhen
yizhou, Chen
Chen, Hua
Pan, Jia
author_facet Xu, Tengye
Sun, Yangting
Shen, Ziju
Chen, Guanqi
Fu, Zhen
yizhou, Chen
Chen, Hua
Pan, Jia
contents Designing reward functions that generalize beyond controlled laboratory settings remains a fundamental challenge in reinforcement learning for robotics. In open-world manipulation problems, a single task can appear in numerous variants through different object instances, positions, and camera viewpoints. Recent vision-based reward models tend to memorize specific pixel distributions and fail to generalize beyond their training conditions. To address this, we propose a framework that learns invariant symbolic reward functions from as few as five demonstrations. The insight is to shift from visual feature-fitting to the discovery of behavioral invariants: task-level properties that remain constant across diverse visual instantiations. The framework has two coupled components: a structural reward formulation that encodes task-level strategies and physical constraints while preserving optimal policy invariance, and a hybrid symbolic-numerical procedure that distills these invariants from demonstrations without online interaction. Experiments on eight Meta-World tasks and three Franka manipulation tasks demonstrate that our method achieves stronger process alignment and policy rollout ranking abilities compared to baselines, accelerating downstream policy learning. Three real-world out-of-distribution experiments further show that the same learned reward generalizes zero-shot to position, viewpoint, and object variations, enabling a single reward representation to be reused across diverse task variants in practice.
format Preprint
id arxiv_https___arxiv_org_abs_2605_22123
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Beyond Pixels: Learning Invariant Rewards for Real-World Robotics From a Few Demonstrations
Xu, Tengye
Sun, Yangting
Shen, Ziju
Chen, Guanqi
Fu, Zhen
yizhou, Chen
Chen, Hua
Pan, Jia
Robotics
Designing reward functions that generalize beyond controlled laboratory settings remains a fundamental challenge in reinforcement learning for robotics. In open-world manipulation problems, a single task can appear in numerous variants through different object instances, positions, and camera viewpoints. Recent vision-based reward models tend to memorize specific pixel distributions and fail to generalize beyond their training conditions. To address this, we propose a framework that learns invariant symbolic reward functions from as few as five demonstrations. The insight is to shift from visual feature-fitting to the discovery of behavioral invariants: task-level properties that remain constant across diverse visual instantiations. The framework has two coupled components: a structural reward formulation that encodes task-level strategies and physical constraints while preserving optimal policy invariance, and a hybrid symbolic-numerical procedure that distills these invariants from demonstrations without online interaction. Experiments on eight Meta-World tasks and three Franka manipulation tasks demonstrate that our method achieves stronger process alignment and policy rollout ranking abilities compared to baselines, accelerating downstream policy learning. Three real-world out-of-distribution experiments further show that the same learned reward generalizes zero-shot to position, viewpoint, and object variations, enabling a single reward representation to be reused across diverse task variants in practice.
title Beyond Pixels: Learning Invariant Rewards for Real-World Robotics From a Few Demonstrations
topic Robotics
url https://arxiv.org/abs/2605.22123