SP
BravenNow
Persona2Web: Benchmarking Personalized Web Agents for Contextual Reasoning with User History
| USA | technology | ✓ Verified - arxiv.org

Persona2Web: Benchmarking Personalized Web Agents for Contextual Reasoning with User History

#Persona2Web #large language models #personalization #user history #ambiguity resolution #web agents #contextual reasoning #benchmarking #cs.CL #cs.AI

📌 Key Takeaways

  • Persona2Web is the first benchmark that measures the effectiveness of personalized web agents on the live web.
  • The benchmark employs the clarify-to-personalize principle, requiring agents to disambiguate queries using implicit user history rather than explicit instructions.
  • It consists of three core components: user histories revealing implicit preferences, ambiguous queries demanding inference, and a reasoning-aware evaluation framework for fine-grained assessment.
  • Extensive experiments across diverse agent architectures, backbone models, history access schemes, and ambiguity levels uncover key challenges in personalization.
  • All code and datasets are publicly released to support reproducibility.

📖 Full Retelling

Who: The research team comprises Serin Kim, Sangam Lee, and Dongha Lee. What: They introduce Persona2Web, a novel benchmark designed to evaluate personalized web agents’ ability to perform contextual reasoning by leveraging user historical data. Where: The benchmark is built on the real, open web, allowing agents to interact with live web content. When: The paper was submitted to arXiv on 19 February 2026. Why: Current large language model–powered web agents lack robust personalization; users rarely provide exhaustive intent details, so agents must resolve ambiguity through inferred preferences and context.

🏷️ Themes

Personalized web agents, User history modeling, Ambiguity resolution in natural language queries, Contextual reasoning, Benchmarking AI systems on the live web, Evaluation frameworks for AI personalization

Entity Intersection Graph

No entity connections available yet for this article.

Deep Analysis

Why It Matters

Persona2Web introduces the first benchmark for evaluating personalized web agents on the open web, highlighting the gap in current agents' ability to infer user preferences from history. This benchmark enables systematic assessment of personalization, guiding future research toward more context-aware web assistants.

Context & Background

  • Large language models have improved web agents but lack personalization
  • Users rarely provide explicit intent, requiring inference from history
  • Persona2Web provides user histories, ambiguous queries, and an evaluation framework
  • Dataset and code are publicly available
  • Benchmark tests various agent architectures and ambiguity levels

What Happens Next

Researchers will use Persona2Web to benchmark and refine personalized web agents, potentially leading to more accurate and user-friendly assistants. The public dataset may spur new models that better handle ambiguity and long-term context.

Frequently Asked Questions

What is Persona2Web?

A benchmark dataset for evaluating personalized web agents on the open web.

How does it differ from existing benchmarks?

It focuses on real web queries, user histories, and ambiguity rather than synthetic tasks.

Where can I access the data?

The code and dataset are publicly available at the URL provided in the paper.

What are the main challenges identified?

Inferring implicit preferences from long histories and resolving ambiguous queries.

}
Original Source
--> Computer Science > Computation and Language arXiv:2602.17003 [Submitted on 19 Feb 2026] Title: Persona2Web: Benchmarking Personalized Web Agents for Contextual Reasoning with User History Authors: Serin Kim , Sangam Lee , Dongha Lee View a PDF of the paper titled Persona2Web: Benchmarking Personalized Web Agents for Contextual Reasoning with User History, by Serin Kim and 2 other authors View PDF HTML Abstract: Large language models have advanced web agents, yet current agents lack personalization capabilities. Since users rarely specify every detail of their intent, practical web agents must be able to interpret ambiguous queries by inferring user preferences and contexts. To address this challenge, we present Persona2Web, the first benchmark for evaluating personalized web agents on the real open web, built upon the clarify-to-personalize principle, which requires agents to resolve ambiguity based on user history rather than relying on explicit instructions. Persona2Web consists of: (1) user histories that reveal preferences implicitly over long time spans, (2) ambiguous queries that require agents to infer implicit user preferences, and (3) a reasoning-aware evaluation framework that enables fine-grained assessment of personalization. We conduct extensive experiments across various agent architectures, backbone models, history access schemes, and queries with varying ambiguity levels, revealing key challenges in personalized web agent behavior. For reproducibility, our codes and datasets are publicly available at this https URL . Subjects: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI) Cite as: arXiv:2602.17003 [cs.CL] (or arXiv:2602.17003v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2602.17003 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Serin Kim [ view email ] [v1] Thu, 19 Feb 2026 01:54:26 UTC (3,104 KB) Full-text links: Access Paper: View a PDF of the paper titled Pers...
Read full article at source

Source

arxiv.org

More from USA

News from Other Countries

🇬🇧 United Kingdom

🇺🇦 Ukraine