Emma Finn

Emma Lucia Byrnes Finn

Oxford University, Harvard University

About Me

I’m an incoming DPhil Student in Statistics at Oxford University, advised by Prof. Yee Whye Teh and Prof. Patrick Rebeschinim supported by a Rhodes Scholarship. I graduated in May of 2026 from Harvard, where I studied Mathematics and Classics for my undergrad degree, with an A.M. in Statistics through the concurrent AB/AM program. My work spans probability theory, statistical modeling, and machine learning—especially interpretable models, stochastic processes, and the mathematics of diffusion. I'm committed to understanding generative AI in order to make it safer, fairer, and more equitable.

While at Harvard, I wrote two senior theses, one in mathematics exploring transport-entropy inequalities and one in Classics exploring ancient mathematical thought in Greek historiography.



Publications, Blogs, and Resources

The Two Clocks and the Innovation Window: When and How Generative Models Learn Rules

Binxu Wang, Emma Finn, Bingbin Liu

An expanded version of the workshop paper, When Rule Learning Breaks: Diffusion Fails to Learn Parity of Many Bits. Under review at NeurIPS 2026.

Read on arXiv

When Rule Learning Breaks: Diffusion Fails to Learn Parity of Many Bits

Binxu Wang, Emma Finn, Bingbin Liu

Workshop paper showing failure modes of diffusion models on parity-like structure. Accepted (oral) to NeurIPS Workshop on Structured Probabilistic Inference & Generative Modeling.

Download PDF

Where the Score Lives: A Wavelet View of Diffusion

Emma Finn, Binxu Wang, T.A. Keller, Demba Ba

Workshop paper introducing a wavelet-based lens on score structure in diffusion models, accepted to NeurIPS Workshop on Structured Probabilistic Inference & Generative Modeling.

Download PDF

Origins of Creativity in Attention-Based Diffusion Models

Emma Finn, T. A. Keller, Manos Theodosis, Demba E. Ba.

ICML 2025 Workshop on High-Dimensional Learning Dynamics. Explores how attention-enhanced score matching uncovers the mechanisms behind creativity in diffusion models.

Read on arXiv

From Here to There: Transport Entropy Inequalities

Emma Finn, supervised by Dr. Mark Sellke and Dr. Kevin Yang · Math Thesis (2025–26)

Download PDF

Learning Artistic Signatures: Symmetry Discovery and Style Transfer

Emma Finn, T. A. Keller, Manos Theodosis, Demba E. Ba

We propose and validate a mathematically grounded definition of artistic style as global symmetries over local textures.

Read on arXiv

Rader's Algorithm for the Fast Fourier Transform

Emma Finn

Junior Paper in Mathematics. Expository paper on the abstract algebra behind the FFT.

Rader's Fast Fourier Transform

Score and Structure

Emma Finn

A blog exploring the intersections of algebraic structures and statistical inference, with visual examples and applications.

Read the Blog

Quantifying the Past: Empirical Tropes in Greek Historiography

Emma Finn · Classics Thesis (2025–26)

A study of how Herodotus, Thucydides, and Xenophon employ quantification as a rhetorical device within their historical narratives.

Download PDF

Phaedrus and Fictionality

Emma Finn

Junior paper in Classics connecting Plato's Phaedrus to how technology interacts with fictionality and irony.

Download PDF

Ruth Hale's Fight for Her Name

Emma Finn, August 13, 2021

Ruth Hale’s legal and cultural battle to keep her name, a landmark fight for women’s identity.

Read here

Education

  • DPhil in Statistics, Oxford University, (intended) May 2030
  • A.M. in Statistics, Harvard University, May 2026
  • A.B. in Mathematics and Classics, Harvard University, May 2026

Teaching

  • Fall 2025: Statistics 210 – Graduate Probability I (TF)
  • Spring 2025: Statistics 111 – Introduction to Statistical Inference (TF)
  • Fall 2024: Statistics 110 – Introduction to Probability (TF)
  • Spring 2024: Math 21B – Linear Algebra and Differential Equations (CA)
  • Fall 2023: Math 21A – Multivariable Calculus (CA)

Other

I love to read, bake (mostly scones and foccacia), and run (slowly!). I also a love doing the NYT-Connections, reading, and horseback riding!


The webpage template is taken from onpix