PersonaTrail: Benchmarking Personalized Web Agents through Browsing Trails

2026-05-30

AI Summary

Existing benchmarks struggle to evaluate web agents’ ability to infer contextual information and user preferences from authentic browsing histories when user instructions are ambiguous. To address this gap, this work introduces the first personalized evaluation benchmark grounded in real user browsing trajectories and proposes the Preference-Aware Contextual Memory framework (PACMem). PACMem decouples browsing history into a dual-channel structured representation comprising factual memory and preference memory. By integrating large language models with a memory retrieval mechanism, the framework effectively models and leverages historical user behavior. Experimental results demonstrate that PACMem significantly outperforms existing memory-based baselines on personalized web navigation tasks, confirming its effectiveness and state-of-the-art performance.

Abstract

Recent advances in large language models have enabled web agents to autonomously execute complex tasks. In practice, users frequently provide underspecified instructions, requiring agents to infer the missing context from their raw browsing histories. Existing benchmarks fail to capture this form of personalization, as they either restrict tasks to fully explicit prompts or abstract web interaction history into simplified forms. To bridge this gap, we introduce PersonaTrail, a benchmark for personalized web agents operating in a managed open web environment. By leveraging realistic browsing trajectories as user history, PersonaTrail evaluates an agent's ability to infer user preferences and recall information from past browsing sessions. We further propose Preference-Aware Contextual Memory (PACMem), a framework that decomposes raw browsing histories into two types of structured memory: factual memories that summarize individual sessions and preference memories that distill recurring behavioral patterns. At inference time, the agent retrieves the most relevant entries from these memories to guide personalized navigation. Extensive experiments show that PACMem consistently outperforms existing memory-based baselines on both tasks.

Problem

• personalized web agents
• browsing trails
• user preferences
• underspecified instructions
• web interaction history

Innovation

• personalized web agents
• browsing trails
• contextual memory
• preference inference
• PACMem