Skyler Hallinan 🤖🧑‍💻

Skyler

I am a Ph.D. student at the University of Southern California advised by Sai Praneeth Karimireddy and Xiang Ren. Previously, I completed a B.S./M.S. in computer science at the University of Washington, where I worked with Yejin Choi. I've also spent time at Microsoft, Samaya AI, Apple, and Amazon AWS.

Links: [Google Scholar] [Twitter] [Github] [CV] [Bio] [Semantic Scholar]

Email: skyler.r.hallinan [at] gmail.com

Research

I aim to build autonomous AI systems capable of complex, long-horizon reasoning, even in unfamiliar environments. I achieve this by shaping their reasoning process: understanding why they fail, then curating better training experiences and inference strategies. Specifically, I focus on:

  • Data Attribution: I build tools to understand how training data shapes model behavior: detecting what models have learned (CoLM 2025) and tracing how it appears in their outputs (ICLR 2025).
  • Training Experience Curation: I optimize the data models learn from at training time, synthesizing and curating data offline, and designing environments and exploration strategies online (NeurIPS 2025).
  • Inference-Time Scaffolding: I design algorithmic structures that transform fixed models into autonomous reasoners, enabling them to refine their outputs (NeurIPS 2023), explore their environment (preprint), and coordinate with tools and agents.

Selected Publications

(for the full list of my publications, please see my Google Scholar)

Selected All

Invited Talks

  • 2025 - "The Surprising Effectiveness of Membership Inference with Simple N-Gram Metrics"
    Google, Mountain View, CA (Virtual) [Video]
  • 2025 - "OpaqueToolsBench: Learning Nuances of Tool Behavior Through Interaction"
    Microsoft, Redmond, WA (Virtual)
  • 2025 - "The Surprising Effectiveness of Membership Inference with Simple N-Gram Metrics"
    USC ISI Seminar [Video]
  • 2024 - "Small but Mighty: Empowering Small Language Models to Outperform Their Larger Counterparts"
    Qualcomm, San Diego, CA (Joint Talk with Jillian Fisher)
  • 2023 - "Enhancing the Utility of Large Language Models with Algorithms"
    Apple, Seattle, WA (Virtual)

Media Coverage

  • 2024 - "AI writing is improving, but it still can’t match human creativity"
    Science [Article]
  • 2021 - "“Every single one of you has what it takes to do great things”: A tribute to the Allen School Class of 2021"
    Paul G. Allen School of Computer Science & Engineering [Article]
  • 2020 - "Six Allen School undergraduates recognized for excellence in research"
    Paul G. Allen School of Computer Science & Engineering [Article]