Saved in:
Bibliographic Details
Main Authors: Schwarzer, Will, Schneider, Jordan, Thomas, Philip S., Niekum, Scott
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.18447
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916629916418048
author Schwarzer, Will
Schneider, Jordan
Thomas, Philip S.
Niekum, Scott
author_facet Schwarzer, Will
Schneider, Jordan
Thomas, Philip S.
Niekum, Scott
contents Existing approaches to reward inference from behavior typically assume that humans provide demonstrations according to specific models of behavior. However, humans often indicate their goals through a wide range of behaviors, from actions that are suboptimal due to poor planning or execution to behaviors which are intended to communicate goals rather than achieve them. We propose that supervised learning offers a unified framework to infer reward functions from any class of behavior, and show that such an approach is asymptotically Bayes-optimal under mild assumptions. Experiments on simulated robotic manipulation tasks show that our method can efficiently infer rewards from a wide variety of arbitrarily suboptimal demonstrations.
format Preprint
id arxiv_https___arxiv_org_abs_2502_18447
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Supervised Reward Inference
Schwarzer, Will
Schneider, Jordan
Thomas, Philip S.
Niekum, Scott
Machine Learning
Existing approaches to reward inference from behavior typically assume that humans provide demonstrations according to specific models of behavior. However, humans often indicate their goals through a wide range of behaviors, from actions that are suboptimal due to poor planning or execution to behaviors which are intended to communicate goals rather than achieve them. We propose that supervised learning offers a unified framework to infer reward functions from any class of behavior, and show that such an approach is asymptotically Bayes-optimal under mild assumptions. Experiments on simulated robotic manipulation tasks show that our method can efficiently infer rewards from a wide variety of arbitrarily suboptimal demonstrations.
title Supervised Reward Inference
topic Machine Learning
url https://arxiv.org/abs/2502.18447