LASER: LLM Agent with State-Space Exploration for Web Navigation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ma, Kaixin, Zhang, Hongming, Wang, Hongwei, Pan, Xiaoman, Yu, Wenhao, Yu, Dong
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909114058145792
author Ma, Kaixin
Zhang, Hongming
Wang, Hongwei
Pan, Xiaoman
Yu, Wenhao
Yu, Dong
author_facet Ma, Kaixin
Zhang, Hongming
Wang, Hongwei
Pan, Xiaoman
Yu, Wenhao
Yu, Dong
contents Large language models (LLMs) have been successfully adapted for interactive decision-making tasks like web navigation. While achieving decent performance, previous methods implicitly assume a forward-only execution mode for the model, where they only provide oracle trajectories as in-context examples to guide the model on how to reason in the environment. Consequently, the model could not handle more challenging scenarios not covered in the in-context examples, e.g., mistakes, leading to sub-optimal performance. To address this issue, we propose to model the interactive task as state space exploration, where the LLM agent transitions among a pre-defined set of states by performing actions to complete the task. This formulation enables flexible backtracking, allowing the model to recover from errors easily. We evaluate our proposed LLM Agent with State-Space ExploRation (LASER) on both the WebShop task and amazon.com. Experimental results show that LASER significantly outperforms previous methods and closes the gap with human performance on the web navigation task.
format Preprint
id arxiv_https___arxiv_org_abs_2309_08172
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle LASER: LLM Agent with State-Space Exploration for Web Navigation
Ma, Kaixin
Zhang, Hongming
Wang, Hongwei
Pan, Xiaoman
Yu, Wenhao
Yu, Dong
Computation and Language
Large language models (LLMs) have been successfully adapted for interactive decision-making tasks like web navigation. While achieving decent performance, previous methods implicitly assume a forward-only execution mode for the model, where they only provide oracle trajectories as in-context examples to guide the model on how to reason in the environment. Consequently, the model could not handle more challenging scenarios not covered in the in-context examples, e.g., mistakes, leading to sub-optimal performance. To address this issue, we propose to model the interactive task as state space exploration, where the LLM agent transitions among a pre-defined set of states by performing actions to complete the task. This formulation enables flexible backtracking, allowing the model to recover from errors easily. We evaluate our proposed LLM Agent with State-Space ExploRation (LASER) on both the WebShop task and amazon.com. Experimental results show that LASER significantly outperforms previous methods and closes the gap with human performance on the web navigation task.
title LASER: LLM Agent with State-Space Exploration for Web Navigation
topic Computation and Language
url https://arxiv.org/abs/2309.08172