Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zhao, Qi, Fu, Haotian, Sun, Chen, Konidaris, George
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2408.16090
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866912057094307840
author Zhao, Qi
Fu, Haotian
Sun, Chen
Konidaris, George
author_facet Zhao, Qi
Fu, Haotian
Sun, Chen
Konidaris, George
contents Long-horizon decision-making tasks present significant challenges for LLM-based agents due to the need for extensive planning over multiple steps. In this paper, we propose a hierarchical framework that decomposes complex tasks into manageable subgoals, utilizing separate LLMs for subgoal prediction and low-level action generation. To address the challenge of creating training signals for unannotated datasets, we develop a reward model that leverages multimodal environment feedback to automatically generate reward signals. We introduce Environment Preference Optimization (EPO), a novel method that generates preference signals from the environment's feedback and uses them to train LLM-based agents. Extensive experiments on ALFRED demonstrate the state-of-the-art performance of our framework, achieving first place on the ALFRED public leaderboard and showcasing its potential to improve long-horizon decision-making in diverse environments.
format Preprint
id arxiv_https___arxiv_org_abs_2408_16090
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle EPO: Hierarchical LLM Agents with Environment Preference Optimization
Zhao, Qi
Fu, Haotian
Sun, Chen
Konidaris, George
Machine Learning
Long-horizon decision-making tasks present significant challenges for LLM-based agents due to the need for extensive planning over multiple steps. In this paper, we propose a hierarchical framework that decomposes complex tasks into manageable subgoals, utilizing separate LLMs for subgoal prediction and low-level action generation. To address the challenge of creating training signals for unannotated datasets, we develop a reward model that leverages multimodal environment feedback to automatically generate reward signals. We introduce Environment Preference Optimization (EPO), a novel method that generates preference signals from the environment's feedback and uses them to train LLM-based agents. Extensive experiments on ALFRED demonstrate the state-of-the-art performance of our framework, achieving first place on the ALFRED public leaderboard and showcasing its potential to improve long-horizon decision-making in diverse environments.
title EPO: Hierarchical LLM Agents with Environment Preference Optimization
topic Machine Learning
url https://arxiv.org/abs/2408.16090