RoboArena
Pranav Atreya*1, Karl Pertsch*1,2, Tony Lee*2, Moo Jin Kim2, Arhan Jain3, Artur Kuramshin4, Clemens Eppner5, Cyrus Neary4, Edward Hu6, Fabio Ramos5, Jonathan Tremblay5, Kanav Arora3, Kirsty Ellis4, Luca Macesanu7, Marcel Torne Villasevil2, Matthew Leonard6, Meedeum Cho8, Ozgur Aslan4, Shivin Dass7, Jie Wang6, William Reger7, Xingfang Yuan6, Xuning Yang5, Abhishek Gupta3, Dinesh Jayaraman6, Glen Berseth4, Kostas Daniilidis6, Roberto Martin-Martin7, Youngwoon Lee8, Percy Liang2, Chelsea Finn2, Sergey Levine1 *: Core contributors RoboArena is a community-run, real-world benchmark for evaluating generalist robot policies, instantiated currently on the DROID robot platform. You can submit your policy to be evaluated by our pool of volunteer evaluators, or become an evaluator yourself. RoboArena evaluations differ from traditional robot evaluation in one key principle: instead of standardizing the set of tasks for which all policies are evaluated, we allow evaluators to pick any environment