VibeMathedMath problems solved by AI

Lower Bounds for Multivariate Independence Polynomials

The multivariate independence polynomial is the partition function of the hard-core model with per-vertex fugacities. The paper proves a lower bound extending to the multivariate setting a result Tao proved in the univariate case, and settles a conjectured generalization for a multiaffine version of the semiproper colouring partition function with two proper colours.

Result
Proved
Status
Resolved
AI contribution
AI co-developed
Method
Argument
Field
Extremal combinatorics; statistical physics
Posed by
Year posed
Years open
Solved
2026-08-05
Model
Aletheia (Gemini Deep Think), Harmonic Aristotle
Vendor
Google DeepMind, Harmonic
Collaborators
Joonkyung Lee, Jaehyeon Seo
Verification
Lean-checked, statement unaudited
Publication
Preprint
Significance
12 / 100
Disclosed cost
Wikipedia
No dedicated article

What the AI did

"The key steps in both proofs were obtained, at least in part, by using Aletheia, a mathematics research agent built upon Gemini Deep Think at Google DeepMind." The paper marks where the agent contributed, publishes raw prompts and outputs in a repository, and calls the work a benchmark demonstrating that current models can in part assist with mathematical research. Theorem 1.4 was separately formalized in Lean 4 with Harmonic Aristotle.

Verification

Theorem 1.4 is formalized in Lean 4 using Harmonic Aristotle, with the files public; the rest of the paper is not formalized, and nobody independent has audited the informal-to-formal correspondence.

Source

arXiv

Discussion