Vision-Based Agent Architectures for Autonomous Enterprise Web Navigation
Архитектуры агентов на основе визуального восприятия для автономной навигации в корпоративной веб-среде
2025-01-01
SCID: 54.1/c6mfmn97
Discuss with AI
enterprise web navigationlanguage groundinglarge language modelsvision-based agentsvisual reasoning
Figures from the paper
Abstract (AI)
The autonomous enterprise web navigation, in general, has been changed along with the introduction of vision-based agent architecture.The visual perception agents are almost replacing the previous variations of the rule-based and DOM-parsing strategies as well as are quite comparable to the human inclinations to the web world interaction.The recent developments in the design and application of such agents and visual-first models, integration principles of large language models, and design principles have also been critically analyzed in the paper.The review also elaborates how language grounding and visual reasoning can be used to achieve agent robustness, adaptability, and generalization to adapt enterprise situations based on the related areas of interest such as autonomous UAV systems, embodied AI, and multimodal robotics.Such issues like effectiveness of inference, error recovery, and ethical aspects are also discussed as the most topical ones, and the new tendencies defining the future generation of the web automation systems also exist.This kind of synthesis will be of value to the academicians, the practices that are currently making efforts on coming up with intelligent agents that can be applied to aid the real world web interfaces within an enterprise. IndexTerms -vision-based agents
Key Findings
1
Insights from autonomous UAVs, embodied AI, and multimodal robotics inform emerging architectures and future directions for web-navigation agents.
2
Large language model integration, language grounding, and visual reasoning are identified as core design principles for improving agent robustness, adaptability, and generalization.
3
The review highlights inference effectiveness, error recovery, and ethical considerations as central challenges for enterprise web automation.
4
Vision-based agents are increasingly replacing rule-based and DOM-parsing approaches for autonomous enterprise web navigation.
5
Visual perception agents are described as more human-like in web interaction and broadly comparable to human interaction tendencies.
Research Object
vision-based agents for autonomous enterprise web navigation
Research Subject
agent architectures, visual perception, language grounding, visual reasoning, robustness, adaptability, generalization, inference effectiveness, error recovery, and ethical aspects in enterprise web automation
Publication Details
Publication Date
2025-01-01
Journal
Publisher
ISSN
Cited by
0
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest