Recent advances in large language models have highlighted their potential for personalized recommendation, where accurately capturing user preferences remains a key challenge. Leveraging their strong reasoning and generalization capabilities, LLMs offer new opportunities for modeling long-term user behavior. To systematically evaluate this, we introduce ALPBench, a Benchmark for Attribution-level Long-term Personal Behavior Understanding. Unlike item-focused benchmarks, ALPBench predicts user-interested attribute combinations, enabling ground-truth evaluation even for newly introduced items. It models preferences from long-term historical behaviors rather than users' explicitly expressed requests, better reflecting enduring interests. User histories are represented as natural language sequences, allowing interpretable, reasoning-based personalization. ALPBench enables fine-grained evaluation of personalization by focusing on the prediction of attribute combinations task that remains highly challenging for current LLMs due to the need to capture complex interactions among multiple attributes and reason over long-term user behavior sequences.
@article{arxiv.2602.03056,
title = {ALPBench: A Benchmark for Attribution-level Long-term Personal Behavior Understanding},
author = {Lu Ren and Junda She and Xinchen Luo and Tao Wang and Xin Ye and Xu Zhang and Muxuan Wang and Xiao Yang and Chenguang Wang and Fei Xie and Yiwei Zhou and Danjun Wu and Guodong Zhang and Yifei Hu and Guoying Zheng and Shujie Yang and Xingmei Wang and Shiyao Wang and Yukun Zhou and Fan Yang and Size Li and Kuo Cai and Qiang Luo and Ruiming Tang and Han Li and Kun Gai},
journal= {arXiv preprint arXiv:2602.03056},
year = {2026}
}