Cognitive Duality for Adaptive Web Agents

TL;DR

This paper introduces CogniWeb, a web agent architecture inspired by human dual-process cognition, integrating offline and online learning to improve efficiency and performance in complex web navigation tasks.

Contribution

It presents a novel dual-process framework for web agents, bridging reactive and deliberative methods, and demonstrates its effectiveness in a modular architecture.

Findings

Achieves 43.96% success rate on WebArena

Reduces token usage by 75%

Provides a unified dual-process perspective for web agents

Abstract

Web navigation represents a critical and challenging domain for evaluating artificial general intelligence (AGI), demanding complex decision-making within high-entropy, dynamic environments with combinatorially explosive action spaces. Current approaches to building autonomous web agents either focus on offline imitation learning or online exploration, but rarely integrate both paradigms effectively. Inspired by the dual-process theory of human cognition, we derive a principled decomposition into fast System 1 and slow System 2 cognitive processes. This decomposition provides a unifying perspective on existing web agent methodologies, bridging the gap between offline learning of intuitive reactive behaviors and online acquisition of deliberative planning capabilities. We implement this framework in CogniWeb, a modular agent architecture that adaptively toggles between fast intuitive processing and deliberate reasoning based on task complexity. Our evaluation on WebArena demonstrates that CogniWeb achieves competitive performance (43.96% success rate) while maintaining significantly higher efficiency (75% reduction in token usage).