Saved in:
Bibliographic Details
Main Authors: Gur, Izzeddin, Furuta, Hiroki, Huang, Austin, Safdari, Mustafa, Matsuo, Yutaka, Eck, Douglas, Faust, Aleksandra
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2307.12856
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911783322648576
author Gur, Izzeddin
Furuta, Hiroki
Huang, Austin
Safdari, Mustafa
Matsuo, Yutaka
Eck, Douglas
Faust, Aleksandra
author_facet Gur, Izzeddin
Furuta, Hiroki
Huang, Austin
Safdari, Mustafa
Matsuo, Yutaka
Eck, Douglas
Faust, Aleksandra
contents Pre-trained large language models (LLMs) have recently achieved better generalization and sample efficiency in autonomous web automation. However, the performance on real-world websites has still suffered from (1) open domainness, (2) limited context length, and (3) lack of inductive bias on HTML. We introduce WebAgent, an LLM-driven agent that learns from self-experience to complete tasks on real websites following natural language instructions. WebAgent plans ahead by decomposing instructions into canonical sub-instructions, summarizes long HTML documents into task-relevant snippets, and acts on websites via Python programs generated from those. We design WebAgent with Flan-U-PaLM, for grounded code generation, and HTML-T5, new pre-trained LLMs for long HTML documents using local and global attention mechanisms and a mixture of long-span denoising objectives, for planning and summarization. We empirically demonstrate that our modular recipe improves the success on real websites by over 50%, and that HTML-T5 is the best model to solve various HTML understanding tasks; achieving 18.7% higher success rate than the prior method on MiniWoB web automation benchmark, and SoTA performance on Mind2Web, an offline task planning evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2307_12856
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle A Real-World WebAgent with Planning, Long Context Understanding, and Program Synthesis
Gur, Izzeddin
Furuta, Hiroki
Huang, Austin
Safdari, Mustafa
Matsuo, Yutaka
Eck, Douglas
Faust, Aleksandra
Machine Learning
Artificial Intelligence
Computation and Language
Pre-trained large language models (LLMs) have recently achieved better generalization and sample efficiency in autonomous web automation. However, the performance on real-world websites has still suffered from (1) open domainness, (2) limited context length, and (3) lack of inductive bias on HTML. We introduce WebAgent, an LLM-driven agent that learns from self-experience to complete tasks on real websites following natural language instructions. WebAgent plans ahead by decomposing instructions into canonical sub-instructions, summarizes long HTML documents into task-relevant snippets, and acts on websites via Python programs generated from those. We design WebAgent with Flan-U-PaLM, for grounded code generation, and HTML-T5, new pre-trained LLMs for long HTML documents using local and global attention mechanisms and a mixture of long-span denoising objectives, for planning and summarization. We empirically demonstrate that our modular recipe improves the success on real websites by over 50%, and that HTML-T5 is the best model to solve various HTML understanding tasks; achieving 18.7% higher success rate than the prior method on MiniWoB web automation benchmark, and SoTA performance on Mind2Web, an offline task planning evaluation.
title A Real-World WebAgent with Planning, Long Context Understanding, and Program Synthesis
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2307.12856