Publications

An LLM–CAS framework for proving asymptotic inequalities

Authors: Ayush Khaitan, Vijay Ganesh
Submitted, 2025  •  arXiv:2510.12350
We build an LLM+CAS framework that proves research‑level asymptotic inequalities. Using it, we resolve multiple questions from MathOverflow and other research forums. This answers a question posed by Terence Tao. We are grateful for Terence's support for our project, including his highlighting of our tool here.

Elementary symmetric polynomials under the fixed point measure

Authors: Ayush Khaitan, Bhargav Narayanan, Ishan Mata
Submitted, 2025  •  arXiv:2505.12178
We prove a surprising inhomogeneous inequality for symmetric polynomials, fully confirming a conjecture of Bhargav Narayanan. The result advances the program of obtaining sharp unconditional lower bounds on the permanent of a matrix.

Computing renormalized curvature integrals on Poincaré–Einstein manifolds

Submitted, 2024  •  arXiv:2404.11319
We describe a general procedure for constructing renormalized curvature integrals on Poincaré–Einstein manifolds.

Ambient metric for manifolds with density and the Ricci flow

Authors: Ayush Khaitan
Advances in Mathematics (accepted), 2023  •  arXiv:2308.02061
We construct fully non‑linear analogues of Perelman’s W‑functional that are monotone along the Ricci flow, and stationary only for shrinking Ricci solitons. We do so by constructing a Fefferman–Graham ambient metric for manifolds with density.

The weighted ambient metric

Authors: Jeffrey S. Case, Ayush Khaitan
SIGMA 18 (2022), 086, 21 pages  •  Journal link
We construct the Fefferman–Graham ambient metric for smooth metric measure spaces.

GJMS operators of smooth metric measure spaces

Authors: Ayush Khaitan
Submitted, 2022  •  arXiv:2203.04719
We construct GJMS operators for smooth metric measure spaces, and prove several properties.

Weighted renormalized volume coefficients

Authors: Ayush Khaitan
Submitted, 2022  •  arXiv:2205.06018
We construct renormalized volume coefficients for smooth metric measure spaces and prove several key properties.