VibeMathedMath problems solved by AI

Lower Bounds for Lebesgue Constants and an Erdős-Turán Interpolation Problem

Localizing Bernstein theory to prove lower bounds for the Lebesgue constants of Lagrange interpolation, with application to a problem of Erdős and Turán and to a conjectured bound from the interpolation literature.

Result
Proved
Status
Resolved
AI contribution
AI co-developed
Method
Argument
Field
Approximation theory
Posed by
Paul Erdős, Pál Turán (interpolation problem)
Year posed
1937
Years open
89y
Solved
2026-03-23
Model
AlphaEvolve, ChatGPT
Vendor
Google DeepMind, OpenAI
Collaborators
Terence Tao
Verification
Unreviewed
Publication
Preprint
Significance
18 / 100
Disclosed cost
Wikipedia
No dedicated article

What the AI did

"After some experimentation using the tool AlphaEvolve, the author was led to conjecture a proof of (1.29) by separately lower bounding each of these two factors. The first of these conjectures was then proven by ChatGPT, and the author was able to prove the second, thus giving a complete proof." Figures generated by Gemini.

Source

arXiv

Discussion