PeopleofSuperintelligence

Researcher

Lionel Levine

Lionel Levine is a Cornell mathematician whose work connects mathematical research with AI evaluation and safety. He earned an A.B. in mathematics at Harvard in 2002 and a Ph.D. in mathematics at UC Berkeley in 2007, then studied random growth and sandpile models. He coauthored FrontierMath in 2024, a benchmark for advanced mathematical reasoning in AI, and EigenBench in 2025, a way to compare language models’ value alignment. As of October 2026, he is a professor of mathematics at Cornell and develops mathematical tools for AI safety.

Timeline

  1. 2002–2007

    Harvard degree and Berkeley doctorate

    He earns an A.B. in mathematics from Harvard in 2002 and a Ph.D. in mathematics from the University of California, Berkeley in 2007. His Berkeley thesis, advised by Yuval Peres, uses free-boundary methods to prove limit shapes for random and deterministic growth models.

  2. 2010

    Logarithmic fluctuations in internal DLA

    With David Jerison and Scott Sheffield, he coauthors a 2010 paper on internal diffusion-limited aggregation. The paper shows that the growing cluster stays within a logarithmic distance of a disk.

  3. 2012

    Fractal solutions in the Abelian sandpile

    With Wesley Pegden and Charles K. Smart, he coauthors a 2012 paper that proves the sandpile PDE has certain fractal solutions. The work gives a mathematical account of fractal patterns in the Abelian sandpile.

  4. 2014

    A Sloan Research Fellowship

    As an assistant professor of mathematics at Cornell, he receives a 2014 Sloan Research Fellowship.

  5. 2021

    A Fellowship from the AMS

    He becomes a 2021 Fellow of the American Mathematical Society for contributions to research and exposition in probability theory and combinatorics.

  6. 2024

    FrontierMath benchmarks AI mathematics

    In 2024, he coauthors FrontierMath, a benchmark of hundreds of original mathematics problems crafted and vetted by mathematicians to assess advanced reasoning in AI. The paper reports that leading AI models solve under 2% of its problems.

  7. 2025

    Comparing language-model values

    In 2025, he coauthors EigenBench, a black-box method for comparatively benchmarking language models’ values. It uses an ensemble of models and a stated value system to score alignment across scenarios.

  8. As of October 2026

    AI safety tools at Cornell

    As of October 2026, he is a professor of mathematics at Cornell. He receives $1.5 million over 18 months from Coefficient Giving to develop mathematical tools for AI safety and evaluate the constitutions that set frontier language models’ characters and values. His 2026 survey invites mathematicians to work on AI safety problems across logic, probability, algebra, and analysis.

Career

  1. NowCornell UniversityAssistant professor of mathematics

Education

  1. 2007University of California, BerkeleyPh.D.
  2. 2002Harvard UniversityA.B.

From the biography.

  1. 11% of AI doctorate holders from 2010 to 2016 founded or led a company or lab by 10 years after graduationIncludes Cornell University. We estimate how often AI doctorate holders founded or led a company or lab by 10 years after graduation, how waiting times compare between founders and top jobs, and which universities trained the most.
  2. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Cornell University. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  3. 84% of people shaping AI with recorded undergraduate and doctoral schools changed institutions for their doctorateIncludes Cornell University. 84% of people shaping AI with recorded undergraduate and doctoral schools changed institutions for their doctorate. Across 2,116 people with both an undergraduate degree and a doctorate recorded, we trace which undergraduate schools feed which doctoral programs, how often students stay at the same school, and how paths differ by birth-country region.
  4. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Cornell University. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.

Sources and credits

Photo: We have no published photo of Lionel Levine yet. Send us a free-licensed photo.

  1. Candidate Talk: The Limiting Shape of Internal DLA with Multiple Sources microsoft.com. Checked 20 October 2026.
  2. Lionel Levine — Math & AI Safety — Cornell University lionellevine.github.io. Checked 20 October 2026.
  3. Logarithmic fluctuations for internal DLA arxiv.org. Checked 20 October 2026.
  4. Apollonian structure in the Abelian sandpile arxiv.org. Checked 20 October 2026.
  5. Lionel Levine and Charles Smart receive 2014 Sloan Fellowships pi.math.cornell.edu. Checked 20 October 2026.
  6. Lionel Levine selected a Fellow of the American Mathematical Society math.cornell.edu. Checked 20 October 2026.
  7. FrontierMath: A Benchmark for Evaluating Advanced Mathematical Reasoning in AI arxiv.org. Checked 20 October 2026.
  8. EigenBench: A Comparative Behavioral Measure of Value Alignment arxiv.org. Checked 20 October 2026.
  9. Cornell mathematicians win grants to ensure AI safety, research integrity math.cornell.edu. Checked 20 October 2026.
  10. Math for AI safety: an invitation for mathematicians arxiv.org. Checked 20 October 2026.