James Flemings

prof_picture.jpg

Los Angeles, CA

Hello! I am a third-year CS PhD student at the University of Southern California working in the USC Meta research center, advised by Murali Annavaram in the SCIP (Super Computing In Pocket) lab. I’m graciously funded by the NSF Graduate Research Fellowship. My research broadly investigates privacy in language models (LLMs). In particular, I’m interested in (1) principally understanding and measuring the privacy leakage of language models and (2) algorithmically controlling the privacy leakage of language models.

I received my BS in computer science and mathematics at the University of Alaska Anchorage in 2022. While there, I pursued research in a wide range of areas, including monochromatic colorings, charged coupled devices, federated learning, and out of distribution data performance in language models.

News

May 15, 2025 Our paper “Estimating Privacy Leakage of Augmented Contextual Knowledge in Language Models” has been accepted to ACL 2025, Main!
Mar 11, 2025 I will be interning at Google as a student researcher this summer, working in the Federated Learning and Analytics team!
Jul 22, 2024 Gave two tech talks @ LinkedIn and TikTok on our work “Differentially Private Prediction of Large Language Models.”
May 15, 2024 Our paper “Differentially Private Knowledge Distillation via Synthetic Text Generation” has been accepted to ACL 2024, Findings
Mar 13, 2024 Our paper “Differentially Private Next-Token Prediction of Large Language Models” has been accepted to NAACL 2024, Main

Selected Publications

  1. preprint
    Differentially Private In-context Learning via Sampling Few-shot Mixed with Zero-shot Outputs
    James Flemings, HaoSheng Gan , Hongyi Li , Meisam Razaviyayn , and Murali Annavaram
    arxiv preprint, 2025
    arXiv
  2. ACL
    Estimating Privacy Leakage of Augmented Contextual Knowledge in Language Models
    James Flemings, Bo Jiang , Wanrong Zhang , Zafar Takhirov , and Murali Annavaram
    In Proceedings of ACL, 2025
    arXiv Code
  3. ACL
    Differentially Private Knowledge Distillation via Synthetic Text Generation
    James Flemings, and Murali Annavaram
    In Findings of ACL, 2024
    arXiv Code
  4. NAACL
    Differentially Private Next-Token Prediction of Large Language Models
    James Flemings, Meisam Razaviyayn , and Murali Annavaram
    In Proceedings of NAACL, 2024
    arXiv Code