flâneur

Voyager | An Open-Ended Embodied Agent with Large Language Models

voyager.minedojo.org · 1,591 words · saved by 4 readers

We introduce Voyager, the first LLM-powered embodied lifelong learning agent in Minecraft that continuously explores the world, acquires diverse skills, and makes novel discoveries without human intervention. Voyager consists of three key components: 1) an automatic curriculum that maximizes exploration, 2) an ever-growing skill library of executable code for storing and retrieving complex behaviors, and 3) a new iterative prompting mechanism that incorporates environment feedback, execution errors, and self-verification for program improvement. Voyager interacts with GPT-4 via blackbox queries, which bypasses the need for model parameter fine-tuning. The skills developed by Voyager are temporally extended, interpretable, and compositional, which compounds the agent's abilities rapidly and alleviates catastrophic forgetting. Empirically, Voyager shows strong in-context lifelong learning capability and exhibits exceptional proficiency in playing Minecraft. It obtains 3.3x more unique it

Voyager | An Open-Ended Embodied Agent with Large Language Models Voyager: An Open-Ended Embodied Agent with Large Language Models Guanzhi Wang 1 2 , Yuqi Xie 3 , Yunfan Jiang 4* , Ajay Mandlekar 1* , Chaowei Xiao 1 5 , Yuke Zhu 1 3 , Linxi "Jim" Fan 1† , Anima Anandkumar 1 2† 1 NVIDIA, 2 Caltech, 3 UT Austin, 4 Stanford, 5 ASU * Equal contribution † Equal advising Corresponding authors: guanzhi@caltech.edu, dr.jimfan.ai@gmail.com arXiv PDF Code Tweet MineDojo Abstract We introduce Voyager, the first LLM-powered embodied lifelon

saved by

related reading