notesum.ai

Published at November 17

Improving User Experience in Preference-Based Optimization of Reward Functions for Assistive Robots

cs.RO

cs.AI

cs.HC

Released Date: November 17, 2024

Authors: Nathaniel Dennler¹, Zhonghao Shi¹, Stefanos Nikolaidis¹, Maja Matarić¹

Aff.: ¹University of Southern California, Los Angeles CA 90007, USA

Arxiv: http://arxiv.org/abs/2411.11182v1

Refer to caption

	Alignment ( $\uparrow$ )			Regret ( $\downarrow$ )			Quality ( $\uparrow$ )
	$d=8$	$d=16$	$d=32$	$d=8$	$d=16$	$d=32$	$d=8$	$d=16$	$d=32$
IG	.848	.606	.374	.331	1.243	2.115	-.001	-.003	.003
CMA-ES	.834	.691	.488	.479	.995	1.876	.688	.601	.450
CMA-ES-IG	.828	.717	.517	.393	.759	1.453	.746	.673	.527