Elamur Research⌕
☰
home / papers / 2026 / PersonaTrace: Synthesizing Realistic Digital Footprints with LLM Agents

PersonaTrace: Synthesizing Realistic Digital Footprints with LLM Agents

Paper · 2026 · Minjia Wang, Yunfeng Wang, Xiao Ma et al.

Summary

Research on digital footprints, such as emails, messages, calendars and purchase traces, is limited by data scarcity. Public corpora like Enron cover narrow slices of activity and usually only one data type. Real user data is locked behind privacy rules, and seed-based synthesis methods fail because there are no suitable seed corpora. PersonaTrace addresses this with a three-stage LLM-agent pipeline:

  1. A Persona Agent builds a profile from demographic priors, including social ties, routines and life events.
  2. An Event Agent retrieves and adapts seed events, then recursively expands them into an "event forest" of sub-events.
  3. An Artifact Generator Agent writes emails, SMS, calendar entries, reminders and wallet passes, while Critic Agents check them for consistency and realism.

Intrinsically, the resulting corpus is more diverse and closer to real email data than earlier synthetic sets. Extrinsically, open LLMs fine-tuned on it generalize best to four real out-of-distribution tasks. This matters because it provides privacy-safe, multi-artifact training data for personal-assistant and retrieval systems.

Key points

  • Setup and grounding: All agents run on Gemini-1.5-Pro at temperature 0.9. Demographic priors come from the 2022 American Community Survey. The event memory is seeded from PersonaHub and deduplicated with MinHash LSH.

  • Event retrieval and expansion: Each persona gets 100 seed events: 30 by embedding similarity (all-MiniLM-L6-v2), 30 sampled at random, and 40 newly generated. Expansion proceeds breadth-first and stops at 300 nodes.

  • Intrinsic evaluation: On email data, PersonaTrace beats other synthetic sets (e.g., FinePersonas-Email, Synthetic-Satellite-Emails) on every diversity measure:

    • pairwise correlation 0.2093 (lower is better)
    • Remote-Clique 0.7898
    • entropy 2.8305

    Its average email length (≈1438) is also closer to that of real corpora.

  • Extrinsic tasks: Evaluation covers email categorization, email drafting, QA and next-message prediction. An ablation replacing the agents with fixed event types and templates sharply lowers performance (e.g., categorization accuracy drops from 0.2733 to 0.0063).

  • Deployment and access: Deployment in an industrial retrieval product gave +7% absolute Recall@10. Data and code are available only on request, under licenses that prohibit surveillance uses.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Harvard UniversityCambridge · United States112 works · 42,023 citations🏢AppleCupertino · United States20 works · 2,146 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksSociology and Political Science · 577 works

Cite

Cite as Minjia Wang, Yunfeng Wang, Xiao Ma, Dexin Lv, Qifan Guo, Lynn Zheng, Benliang Wang, Lei Wang, Jiannan Li, Yongwei Xing, Junzhe Xu and Zheng Sun. 2026. PersonaTrace: Synthesizing Realistic Digital Footprints with LLM Agents. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 5: Industry Track).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-industry.5

@inproceedings{wang2026personatrace,
  title = {{PersonaTrace: Synthesizing Realistic Digital Footprints with LLM Agents}},
  author = {Minjia Wang, Yunfeng Wang, Xiao Ma, Dexin Lv, Qifan Guo, Lynn Zheng, Benliang Wang, Lei Wang, Jiannan Li, Yongwei Xing, Junzhe Xu and Zheng Sun},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 5: Industry Track)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-industry.5},
  url = {https://doi.org/10.18653/v1/2026.eacl-industry.5}
}

References

41 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Abdulla Al-Subaiey, Mohammed Al-Thani, Naser Abdullah Alam, Kaniz Fatema Antora, Amith Khandakar, and SM Ashfaq Uz Zaman. 2024. Novel interpretable and robust web-based ai platform for phishing email detection. Computers and Electrical Engineering, 120:109625. find
  2. Marco Braga, Pranav Kasela, Alessandro Raganato, and Gabriella Pasi. 2024. Synthetic data generation with large language models for personalized community question answering. arXiv preprint arXiv:2410.22182. \#\# Ethical Considerations Andrei Z Broder, Moses Charikar, Alan M Frieze, and Michael Mitzen… — [arXiv:2410.22182](https://arxiv.org/abs/2410.22182) find
  3. Privacy and data protection. PersonaTrace has been designed with ethical safeguards at its core. Importantly, the framework relies exclusively on Heng Er Metilda Chee, Jiayin Wang, Zhiqiang Guo, Weizhi Ma, Qinglang Guo, and Min Zhang. 2025. A 106k multi-topic multilingual conversational user dataset… — [arXiv:2502.19108](https://arxiv.org/abs/2502.19108) find
  4. Samuel Rhys Cox, Yunlong Wang, Ashraf Abdul, Christian Von Der Weth, and Brian Y. Lim. 2021. Directed diversity: Leveraging language embedding distances for collective creativity in crowd ideation. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems, pages 1–35. find
  5. Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,… find
  6. Leon Fröhling, Gianluca Demartini, and Dennis Assenmacher. 2024. Personas with attitudes: Controlling llms for diverse data annotation. arXiv preprint arXiv:2410.11745. find
  7. Mohammad Khalil, Farhad Vadiee, Ronas Shakya, and Qinyi Liu. 2025. Creating artificial students that never existed: Leveraging large language models and ctgans for synthetic data generation. In Proceedings of the 15th International Learning Analytics and Knowledge Conference, pages 439–450. find
  8. Saumya Gandhi, Ritu Gala, Vijay Viswanathan, Tongshuang Wu, and Graham Neubig. 2024. Better synthetic data by retrieving and transforming existing datasets. arXiv preprint arXiv:2404.14361. find
  9. Bryan Klimt and Yiming Yang. 2004. The enron corpus: A new dataset for email classification research. In European conference on machine learning, pages 217–226. Springer. find
  10. Tao Ge, Xin Chan, Xiaoyang Wang, Dian Yu, Haitao Mi, and Dong Yu. 2025. Scaling synthetic data creation with 1,000,000,000 personas. Preprint, arXiv:2406.20094.
  11. Ayinoluwa Kolawole and Shukurat Rahmon. 2024. Utilizing digital footprint analysis for end-to-end risk-based authentication in medical billing systems. World Journal of Advanced Engineering Technology and Sciences, pages 166–179. find
  12. Scott A Golder and Michael W Macy. 2014. Digital footprints: Opportunities and challenges for online social research. Annual review of sociology, 40(1):129–152. find
  13. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  14. Haoran Li, Qingxiu Dong, Zhengyang Tang, Chaojun Wang, Xingxing Zhang, Haoyang Huang, Shaohan Huang, Xiaolong Huang, Zeqiang Huang, Dongdong Zhang, and 1 others. 2024a. Synthetic data (almost) from scratch: Generalized instruction tuning for language models. arXiv preprint arXiv:2402.13064. find
  15. Francesco Greco, Giuseppe Desolda, Andrea Esposito, Alessandro Carelli, and 1 others. 2024. David versus goliath: Can machine learning detect llm-generated text? a case study in the detection of phishing emails. In The Italian Conference on CyberSecurity. find
  16. Jiayu Li, Xuan Zhu, Fang Liu, and Yanjun Qi. 2024b. Aide: Task-specific fine tuning with attribute guided multi-hop data expansion. arXiv preprint arXiv:2412.06136. — [arXiv:2412.06136](https://arxiv.org/abs/2412.06136) find
  17. Yanran Li, Hui Su, Xiaoyu Shen, Wenjie Li, Ziqiang Cao, and Shuzi Niu. 2017. DailyDialog: A manually labelled multi-turn dialogue dataset. In Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 986–995, Taipei, Taiwan. Asian Federati…
  18. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, and 1 others. 2022. Lora: Low-rank adaptation of large language models. ICLR, 1(2):3.
  19. Yiming Huang, Xiao Liu, Yeyun Gong, Zhibin Gou, Yelong Shen, Nan Duan, and Weizhu Chen. 2025. Key-point-driven data synthesis with its enhancement on mathematical reasoning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 24176–24184. find
  20. Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74–81.
  21. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, and 1 others. 2023. Self-refine: Iterative refinement with self-feedback. Advances in Neural Information Processing Systems, 36:46534–46594. find
  22. Yue Huang, Siyuan Wu, Chujie Gao, Dongping Chen, Qihui Zhang, Yao Wan, Tianyi Zhou, Chaowei Xiao, Jianfeng Gao, Lichao Sun, and 1 others. 2024. Datagen: Unified synthetic dataset generation via large language models. In The Thirteenth International Conference on Learning Representations. find
  23. Leland McInnes, John Healy, and James Melville. 2018. Umap: Uniform manifold approximation and projection for dimension reduction. arXiv preprint arXiv:1802.03426.
  24. Pegah Jandaghi, Xianghai Sheng, Xinyi Bai, Jay Pujara, and Hakim Sidahmed. 2024. Faithful persona-based conversational dataset generation with large language Parisa Mehdi Gholampour and Rakesh M. Verma. 2023. Adversarial robustness of phishing email detection models. In Proceedings of the 9th ACM In… find
  25. Nagagopiraju Vullam, Sai Srinivas Vellela, Venkateswara Reddy, M Venkateswara Rao, Khader Basha SK, and 1 others. 2023. Multi-agent personalized recommendation system in e-commerce based on user. In 2023 2nd International Conference on Applied Artificial Intelligence and Computing (ICAAIC), pages 11… find
  26. Giuseppe Michele Padricelli and Marianna Coppola. 2024. Challenges in digital social research methods: Algorithms, traces and footprints. a resume of the current debate. Italian Sociological Review, 14(10S):515–530. find
  27. Yuxiang Wei, Zhe Wang, Jiawei Liu, Yifeng Ding, and Lingming Zhang. 2023. Magicoder: Empowering code generation with oss-instruct. arXiv preprint arXiv:2312.02120. find
  28. Ashwinee Panda, Christopher A Choquette-Choo, Zhengming Zhang, Yaoqing Yang, and Prateek Mittal. 2024. Teach llms to phish: Stealing private information from language models. arXiv preprint arXiv:2403.00871. — [arXiv:2403.00871](https://arxiv.org/abs/2403.00871) find
  29. Can Xu, Qingfeng Sun, Kai Zheng, Xiubo Geng, Pu Zhao, Jiazhan Feng, Chongyang Tao, and Daxin Jiang. 2023. Wizardlm: Empowering large language models to follow complex instructions. arXiv preprint arXiv:2304.12244.
  30. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, and 25 others. 2025. Qwen2.5 technical report. Preprint, arXiv:2412.15115. find
  31. Zhangchen Xu, Fengqing Jiang, Luyao Niu, Yuntian Deng, Radha Poovendran, Yejin Choi, and Bill Yuchen Lin. 2024. Magpie: Alignment data synthesis from scratch by prompting aligned llms with nothing. arXiv preprint arXiv:2406.08464.
  32. Longhui Yu, Weisen Jiang, Han Shi, Jincheng Yu, Zhengying Liu, Yu Zhang, James T Kwok, Zhenguo Li, Adrian Weller, and Weiyang Liu. 2023. Metamath: Bootstrap your own mathematical questions for large language models. arXiv preprint arXiv:2309.12284.
  33. Kate Shiells, Nina Di Cara, Anya Skatova, Oliver SP Davis, Claire MA Haworth, Andy L Skinner, Richard Thomas, Alastair R Tanner, John Macleod, Nicholas J Timpson, and 1 others. 2022. Participant acceptability of digital footprint data collection strategies: an exemplar approach to participant engage… find
  34. Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. 2018. Personalizing dialogue agents: I have a dog, do you have pets too? arXiv preprint arXiv:1801.07243.
  35. Sathya Krishnan Suresh, Wu Mengjun, Tushar Pranav, and Eng Siong Chng. 2024. Diasynth: Synthetic dialogue generation framework for low resource dialogue applications. arXiv preprint arXiv:2409.19020. find
  36. Shiyue Zhang, Asli Celikyilmaz, Jianfeng Gao, and Mohit Bansal. 2021. Emailsum: Abstractive email thread summarization. arXiv preprint arXiv:2107.14691. find
  37. Juntao Tan, Liangwei Yang, Zuxin Liu, Zhiwei Liu, Rithesh Murthy, Tulika Manoj Awalgaonkar, Jianguo Zhang, Weiran Yao, Ming Zhu, Shirley Kokane, Silvio Savarese, Huan Wang, Caiming Xiong, and Shelby Heinecke. 2025. Personabench: Evaluating ai models on understanding personal information through acce… find
  38. Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q Weinberger, and Yoav Artzi. 2019. Bertscore: Evaluating text generation with bert. arXiv preprint arXiv:1904.09675. find
  39. Yilin Zhao, Xinbin Yuan, Shanghua Gao, Zhijie Lin, Qibin Hou, Jiashi Feng, and Daquan Zhou. 2023. Chatanything: Facetime chat with llm-enhanced personas. arXiv preprint arXiv:2311.06772. — [arXiv:2311.06772](https://arxiv.org/abs/2311.06772) find
  40. Shuo Tang, Xianghe Pang, Zexi Liu, Bohan Tang, Rui Ye, Tian Jin, Xiaowen Dong, Yanfeng Wang, and Siheng Chen. 2024. Synthesizing post-training data for llms through multi-agent simulation. arXiv preprint arXiv:2410.14251.
  41. U.S. Census Bureau. 2022. American community survey 5-year estimates: 2018-2022. https://data.census.gov/. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-personatrace-synthesizing-realistic-digital-footprints/