Yanan Long
I am a machine learning researcher and consulting CTO at StickFlux Labs, working on AI evaluation, Bayesian statistics and participatory AI governance. I study what evaluation results can support, how reporting choices affect conclusions, and how uncertainty and stakeholder perspectives should inform decisions about AI systems.
My recent work includes two ICML 2026 papers on social-impact evaluation reporting and benchmark saturation, the Evaluation Cards and Every Eval Ever preprints, and a Bayesian audit of public frontier-AI evaluation archives. I also work on multilingual mechanistic interpretability and the relationship between technical evidence and participatory governance.
My background is in biomedical machine learning, including drug–drug interaction prediction, wearable time-series analysis and reproducible training pipelines. I received my PhD in Chemistry (Biomedical Data Science) from the University of Chicago, where I also worked as a researcher data scientist.
Since 2023, I have been a core organizer at Queer in AI. Alongside technical research, I use participatory design and critical HCI to examine how communities shape AI evaluation and governance.
Publications · CV · Consulting through StickFlux Labs
Recent updates
August 2026: An updated version of our ICML paper When AI Benchmarks Plateau is available.
July 2026: Two papers at ICML 2026: Who Evaluates AI’s Social Impacts? and When AI Benchmarks Plateau. My paper Bayesian Inference and Decision Audits for Public Archives of Frontier AI Evaluations was also presented at the CTB and HT workshops.
June 2026: New preprint: Every Eval Ever: A Unifying Schema and Community Repository for AI Evaluation Results.
June 2026: New preprint: Evaluation Cards: An Interpretive Layer for AI Evaluation Reporting.
June 2026: Two papers at ACM FAccT 2026: Taking Stock at FAccT and Challenges to Grassroots Organization Engagement with AI Policy, alongside tutorial and CRAFT contributions.
Recommended citation: