Behaviorally Grounded User Profiles from the Wild for Personalized Alignment and Multi-Perspective Reasoning
Summary
This paper introduces a framework for extracting behaviorally grounded user profiles from social media posts to enhance personalized alignment and multi-perspective reasoning in large language models, outperforming synthetic profile methods.
View Cached Full Text
Cached at: 09/02/26, 05:42 AM
# Behaviorally Grounded User Profiles from the Wild for Personalized Alignment and Multi-Perspective Reasoning Source: [https://arxiv.org/abs/2609.00014](https://arxiv.org/abs/2609.00014) [View PDF](https://arxiv.org/pdf/2609.00014) > Abstract:Persona\-driven techniques increasingly adapt large language models \(LLMs\) to diverse contexts\. However, existing methods predominantly rely on rigid, synthetic personas that flatten individual variation, rely on stereotypes, and miss the nuanced signals driving actual human preferences\. We introduce profile behavioral grounding, a framework for extracting open\-ended, high\-fidelity user profiles directly from authentic, anonymized social media posts\. We evaluate these profiles across two paradigms: train\-time personalization via supervised finetuning \(SFT\) and non\-parametric test\-time multi\-perspective reasoning\. Across complex recommendation and open\-ended query benchmarks, behaviorally grounded profiles consistently improve base models and outperform synthetic profile baselines, driving stronger parametric alignment and enabling richer, multifaceted reasoning\. Our findings establish open\-ended, behavior\-derived profiles as a highly diverse and effective foundation for the next generation of personalized language systems\. Our code base is available at[this https URL](https://github.com/ServiceNow/behavior-grounding)\. ## Submission history From: Yuxuan Li \[[view email](https://arxiv.org/show-email/83fb86fe/2609.00014)\] **\[v1\]**Tue, 11 Aug 2026 10:19:11 UTC \(12,494 KB\)
Similar Articles
SocialPersona: Benchmarking Personalized Profiling and Response with Multimodal Social-Media Context
Introduces SocialPersona, a benchmark for evaluating multimodal large language models on their ability to recover revealed preferences from longitudinal social-media timelines and use them in personalized dialogue.
A Persona-Based Evaluation Framework for Pluralistic Alignment in Generative AI
This paper introduces a persona-based evaluation framework that uses synthetic cognitive profiles to represent diverse human perspectives for pluralistic alignment in generative AI, addressing the limitations of monolithic benchmarks.
ProfileFoundry: A Synthetic Person-Object Substrate for Privacy, Memory, and Tool-Use Evaluation in LLM Agent
ProfileFoundry introduces a deterministic synthetic person-object dataset of 100,000 profiles with 700k+ events, designed to evaluate LLM agents on privacy, memory, and tool-use while ensuring inspectability and consistency.
Generating Diverse Personas for User Simulators to Test Interview Dialogue Systems
This paper proposes a method to automatically generate diverse user personas using large language models for testing interview dialogue systems, reducing manual effort and increasing variation in simulated user behaviors.
LUNAR: Benchmarking Personalized Large Language Models on UNiversal User BehAvioR Logs
This paper introduces LUNAR, a benchmark for evaluating how large language models personalize responses from longitudinal app interaction histories across daily-life domains such as clothing, food, housing, and mobility. Experiments on 19 mainstream LLMs reveal that effective personalization depends on evidence selection and cross-domain integration, and that stronger personalization can come at the cost of privacy protection.