StreamProfileBench: A Benchmark for Fine-Grained User Profile Inference in Real-World Streaming Scenarios

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Wang, Sizhe, Duan, Feiyu, Wang, Juelin, Zhang, Liwen, Wei, Zhongyu
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914601828876288
author Wang, Sizhe
Duan, Feiyu
Wang, Juelin
Zhang, Liwen
Wei, Zhongyu
author_facet Wang, Sizhe
Duan, Feiyu
Wang, Juelin
Zhang, Liwen
Wei, Zhongyu
contents Large Language Models (LLMs) have reshaped user profiling, yet current evaluations mainly focus on static data snapshots. This paradigm overlooks the reality of personalized systems, where User-Generated Content (UGC) arrives continuously and fine-grained profile evolve rapidly. To bridge this gap, we introduce StreamProfileBench, a large-scale benchmark for fine-grained streaming user profiling. We formalize streaming user profiling as a continuous state maintenance task and curate a highly authentic dataset comprising over 120,000 UGC posts from 7,000+ real users across five diverse platforms. By leveraging the temporal correlation of user interests, we further propose a novel, annotation-free evaluation framework. Extensive experiments across 14 leading LLMs reveal that continuous profile updating remains an open challenge. Models exhibit a systemic conservative bias, over-retaining past interests while failing to recognize interest decay. Ablation experiments further validate the practical utility and necessity of the streaming paradigm.
format Preprint
id arxiv_https___arxiv_org_abs_2605_25758
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle StreamProfileBench: A Benchmark for Fine-Grained User Profile Inference in Real-World Streaming Scenarios
Wang, Sizhe
Duan, Feiyu
Wang, Juelin
Zhang, Liwen
Wei, Zhongyu
Computation and Language
Large Language Models (LLMs) have reshaped user profiling, yet current evaluations mainly focus on static data snapshots. This paradigm overlooks the reality of personalized systems, where User-Generated Content (UGC) arrives continuously and fine-grained profile evolve rapidly. To bridge this gap, we introduce StreamProfileBench, a large-scale benchmark for fine-grained streaming user profiling. We formalize streaming user profiling as a continuous state maintenance task and curate a highly authentic dataset comprising over 120,000 UGC posts from 7,000+ real users across five diverse platforms. By leveraging the temporal correlation of user interests, we further propose a novel, annotation-free evaluation framework. Extensive experiments across 14 leading LLMs reveal that continuous profile updating remains an open challenge. Models exhibit a systemic conservative bias, over-retaining past interests while failing to recognize interest decay. Ablation experiments further validate the practical utility and necessity of the streaming paradigm.
title StreamProfileBench: A Benchmark for Fine-Grained User Profile Inference in Real-World Streaming Scenarios
topic Computation and Language
url https://arxiv.org/abs/2605.25758