← Back to all news
arXiv Machine Learning By Rongchao Xu, Dahai Yu, Lin Jiang, Guang Wang

ZeroHAT: Behavior-Conditioned Zero-Shot Human Activity Trace Generation

Read the original on arXiv Machine Learning →

ZeroHAT is a new framework for generating synthetic human activity traces (HATs) in a target region without any real data from that region. It transfers behavioral patterns learned from real HATs in source regions and adapts them using publicly available contextual information about the target region. The system includes a consistency-aware intent extractor, a cross-region behavioral cloning module, and a behavior-conditioned activity realization module, and it outperforms the strongest baseline by 4.5–6.4× in downstream utility and improves fidelity by 15.6–40.8% across ten cities.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

Related stories

arXiv AI
Aug 7

LUNAR: Benchmarking Personalized Large Language Models on UNiversal User BehAvioR Logs

arXiv:2608. 05246v1 Announce Type: new Abstract: Existing personalized LLM benchmarks primarily rely on textual personas or isolated behavioral signals, providing limited evaluation of cross-domain behavioral personalization, where responses must be grounded in heterogeneous daily-life activities.

By Jiahao Zhang, Yongzhi Tong, Zelin Fu, Pengde Zhao, Yanmei Jiang, Jiang Feng, Min Yang
llmsefficiencybenchmarkssafety
More like this →
arXiv Machine Learning
Oct 9

MotiveMob: Motivation as Semantic Action for Closed-Loop Human Mobility Generation

MotiveMob is a motivation-driven autoregressive framework for generating human mobility trajectories. It models the "why-where-when" decision process by first predicting a plausible motivation for each movement step and then jointly generating the next location and arrival time. The approach is evaluated under user-level and temporal distribution shifts, including seasonal changes and COVID-19 disruptions, and consistently outperforms pretraining- and prompting-based baselines in distributional fidelity.

By Mengkun Gao, Zengqing Wu, Renhe Jiang, Jiawei Wang, Yusong Wang, Chuang Yang, Shuyuan Zheng, Makoto Onizuka, Chuan Xiao
llms
More like this →