OmniBehavior：首个基于真实数据的大语言模型用户行为模拟评测基准
Towards Real-world Human Behavior Simulation: Benchmarking Large Language Models on Long-horizon, Cross-scenario, Heterogeneous Behavior Traces
研究团队推出 OmniBehavior，这是首个完全基于真实数据构建的预测性用户模拟评测基准，统一整合长时程、跨场景、异质行为模式。实验表明现有 LLM 难以准确预测这类复杂行为，且性能随上下文窗口扩大而趋于停滞。对比真实行为还发现 LLM 收敛于“积极平均人”，呈现过度活跃、人格同质化和乌托邦偏差，丢失个体差异与长尾行为。该工作已被 NeurIPS 2026 接收。
Authors:Jiawei Chen, Ruoxi Xu, Boxi Cao, Ruotong Pan, Yunfei Zhang, Yifei Hu, Yong Du, Tingting Gao, Han Li, Yaojie Lu, Yingfei Sun, Xianpei Han, Le Sun, Xiangyu Wu, Hongyu Lin
Abstract:The emergence of Large Language Models (LLMs) has illuminated the potential for a general-purpose user simulator. However, existing benchmarks remain constrained to isolated scenarios, narrow action spaces, or synthetic data, failing to capture the holistic nature of authentic human behavior. To bridge this gap, we introduce OmniBehavior, the first predictive user simulation benchmark constructed entirely from real-world data, integrating long-horizon, cross-scenario, and heterogeneous behavioral patterns into a unified framework. Based on this benchmark, we first provide empirical evidence that previous datasets with isolated scenarios suffer from tunnel vision, whereas real-world decision-making relies on long-term, cross-scenario causal chains. Extensive evaluations of state-of-the-art LLMs reveal that current models struggle to accurately predict these complex behaviors, with performance plateauing even as context windows expand. Crucially, a systematic comparison between simulated and authentic behaviors uncovers a fundamental structural bias: LLMs tend to converge towards a positive average person, exhibiting hyper-activity, persona homogenization, and a utopian bias. This results in the loss of individual differences and long-tail behaviors, highlighting critical directions for future high-fidelity simulation research.
| Comments: | Accepted by NeurIPS 2026 |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2604.08362 [cs.CL] |
| (or arXiv:2604.08362v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2604.08362 arXiv-issued DOI via DataCite |
Submission history
From: Jiawei Chen [view email]
[v1]
Thu, 9 Apr 2026 15:26:21 UTC (3,581 KB)
[v2]
Thu, 21 May 2026 16:20:11 UTC (3,540 KB)
[v3]
Sat, 24 Oct 2026 10:11:59 UTC (3,540 KB)
来源：arXiv cs.AI · arxiv.org