Ashley Libasci

Future of Math Research

I was selected for Future of Math Research, a competitive faculty-mentored project at the Illinois Mathematics Lab on math research in the AI age. We generate proofs with language models, verify them in Lean, and communicate the results with generative AI.

The question

A proof assistant doesn't care who wrote a proof. If Lean accepts it with no goals left, it's correct, whether a person or a model produced it. That lets us take model output seriously without taking it on trust, and it raises real questions about how research gets done and shared.

SumOdd.lean
open Finset

-- 1 + 3 + 5 + ... + (2n - 1) = n²
theorem sum_odd (n : ℕ) :
    ∑ i ∈ range n, (2 * i + 1) = n ^ 2 := by
  induction n with
  | zero => simp
  | succ k ih => rw [sum_range_succ, ih]; ring

An illustrative example in Lean 4 with Mathlib, not one of our results.

My role

I'm one of the undergraduate researchers on a team with faculty mentor Jacob Beckey. It runs as MATH 492, about ten hours a week.

The Proof of History

Part of the project is a math-history site about how mathematicians have shared and checked their work, with case files, a timeline and sources. It's a work in progress and goes public once the last round of reviewer corrections is in.

The Proof of History home page: a dark green page with the headline How Mathematicians Shared and Tested Their Work.
Work in progress: the opening of The Proof of History.

Project

Program
Illinois Mathematics Lab, UIUC
Course
MATH 492, 3 credits
Term
August 2026 to now
Team
A faculty mentor and 3 to 5 undergraduates
Tools
Lean 4, language models, generative AI