Scholarlink

We propose FLLL \(^{3}\) M—Federated Learning with Large Language Models for Mobility Modeling—a privacy-preserving framework for Next-Location Prediction (NxLP). By retaining user data locally and leveraging LLMs through an efficient outer product mechanism, FLLL \(^{3}\) M ensures high accuracy with low resource demands. It achieves state-of-the-art results on Gowalla (Acc@1: 12.55, MRR: 0.1422), WeePla-ce (10.71, 0.1285), Brightkite (10.42, 0.1169), and FourSquare (8.71, 0.1023), while reducing parameters by up to 45.6% and memory usage by 52.7%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Can We Predict Your Next Move Without Breaking Your Privacy?

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Arpita Soni,
  • Sahil Tripathi,
  • Gautam Siddharth Kashyap,
  • Manaswi Kulahara,
  • Mohammad Anas Azeez,
  • Zohaib Hasan Siddiqui,
  • Nipun Joshi,
  • Jiechao Gao

摘要

We propose FLLL \(^{3}\) M—Federated Learning with Large Language Models for Mobility Modeling—a privacy-preserving framework for Next-Location Prediction (NxLP). By retaining user data locally and leveraging LLMs through an efficient outer product mechanism, FLLL \(^{3}\) M ensures high accuracy with low resource demands. It achieves state-of-the-art results on Gowalla (Acc@1: 12.55, MRR: 0.1422), WeePla-ce (10.71, 0.1285), Brightkite (10.42, 0.1169), and FourSquare (8.71, 0.1023), while reducing parameters by up to 45.6% and memory usage by 52.7%.