Large Language Models as Generalizable Policies for Embodied Tasks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Szot, Andrew, Schwarzer, Max, Agrawal, Harsh, Mazoure, Bogdan, Talbott, Walter, Metcalf, Katherine, Mackraz, Natalie, Hjelm, Devon, Toshev, Alexander
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916207690514432
author Szot, Andrew
Schwarzer, Max
Agrawal, Harsh
Mazoure, Bogdan
Talbott, Walter
Metcalf, Katherine
Mackraz, Natalie
Hjelm, Devon
Toshev, Alexander
author_facet Szot, Andrew
Schwarzer, Max
Agrawal, Harsh
Mazoure, Bogdan
Talbott, Walter
Metcalf, Katherine
Mackraz, Natalie
Hjelm, Devon
Toshev, Alexander
contents We show that large language models (LLMs) can be adapted to be generalizable policies for embodied visual tasks. Our approach, called Large LAnguage model Reinforcement Learning Policy (LLaRP), adapts a pre-trained frozen LLM to take as input text instructions and visual egocentric observations and output actions directly in the environment. Using reinforcement learning, we train LLaRP to see and act solely through environmental interactions. We show that LLaRP is robust to complex paraphrasings of task instructions and can generalize to new tasks that require novel optimal behavior. In particular, on 1,000 unseen tasks it achieves 42% success rate, 1.7x the success rate of other common learned baselines or zero-shot applications of LLMs. Finally, to aid the community in studying language conditioned, massively multi-task, embodied AI problems we release a novel benchmark, Language Rearrangement, consisting of 150,000 training and 1,000 testing tasks for language-conditioned rearrangement. Video examples of LLaRP in unseen Language Rearrangement instructions are at https://llm-rl.github.io.
format Preprint
id arxiv_https___arxiv_org_abs_2310_17722
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Large Language Models as Generalizable Policies for Embodied Tasks
Szot, Andrew
Schwarzer, Max
Agrawal, Harsh
Mazoure, Bogdan
Talbott, Walter
Metcalf, Katherine
Mackraz, Natalie
Hjelm, Devon
Toshev, Alexander
Machine Learning
Artificial Intelligence
Computation and Language
We show that large language models (LLMs) can be adapted to be generalizable policies for embodied visual tasks. Our approach, called Large LAnguage model Reinforcement Learning Policy (LLaRP), adapts a pre-trained frozen LLM to take as input text instructions and visual egocentric observations and output actions directly in the environment. Using reinforcement learning, we train LLaRP to see and act solely through environmental interactions. We show that LLaRP is robust to complex paraphrasings of task instructions and can generalize to new tasks that require novel optimal behavior. In particular, on 1,000 unseen tasks it achieves 42% success rate, 1.7x the success rate of other common learned baselines or zero-shot applications of LLMs. Finally, to aid the community in studying language conditioned, massively multi-task, embodied AI problems we release a novel benchmark, Language Rearrangement, consisting of 150,000 training and 1,000 testing tasks for language-conditioned rearrangement. Video examples of LLaRP in unseen Language Rearrangement instructions are at https://llm-rl.github.io.
title Large Language Models as Generalizable Policies for Embodied Tasks
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2310.17722