VibeMathedMath problems solved with AI

The uniform Turán density of the tetrahedron

The uniform Turán density πu(H)\pi_u(H) of a 3-uniform hypergraph HH is the smallest dd such that every sufficiently large hypergraph in which all linear-sized subsets induce density above dd contains HH. Erdős and Sós asked in the 1982 paper that founded the subject for the value of πu(K4(3))\pi_u(K_4^{(3)}), the complete 3-graph on four vertices, and for πu(K4(3))\pi_u(K_4^{(3)-}). The second was settled at 1/41/4 in 2018 by Glebov, Kráľ and Volec; the tetrahedron itself resisted for forty-four years, with 1/21/2 the conjectured value from a known lower-bound construction. What is πu(K4(3))\pi_u(K_4^{(3)})?

Result
Proved(see note)
Status
Resolved
AI contribution
AI-assisted
Method
Argument
Field
Extremal hypergraph theory
Posed by
Paul Erdős and Vera T. Sós, in the 1982 paper that introduced uniform Turán densities
Year posed
1982
Years open
44y
Solved
2026-09-10
Model
ChatGPT 6 Pro; Aristotle
Vendor
OpenAI; Harmonic
Collaborators
Matija Bucić
Verification
Unreviewed
Publication
Preprint
Significance
48 / 100
Disclosed cost
Wikipedia
No dedicated article

What was actually shown

πu(K4(3))=1/2\pi_u(K_4^{(3)}) = 1/2, matching the long-known lower-bound construction. The AI contribution is one identity inside the argument rather than the argument, which is why this is recorded as AI-assisted; it is included in the catalog because the identity is a step of the proof, not a matter of exposition.

What the AI did

From the paper's declaration of AI use: identity (6) in the proof was found by ChatGPT 6 Pro, after an early draft containing the author's ideas was uploaded to it. Separately, the paper relies on a computational verifier; besides the author hand-checking that verifier's correctness, Aristotle was used to check it in Lean and to run a separate audit. The mathematical frame, the reduction and the write-up are the author's; the model supplied one identity inside it.

Verification

Checked here on 22 September 2026 against arXiv:2609.11802: the abstract states the value 1/21/2 and that it answers a question of Erdős and Sós from the founding 1982 paper, and the introduction names that paper and the companion question about K4(3)K_4^{(3)-}. The AI declaration is quoted as written. The mathematics was not checked here. Aristotle's Lean check covers the computational verifier used inside the proof, not the proof as a whole, so this is not a formalisation of the theorem and the entry does not claim one. Twelve days old, no referee.

Sources

Changelog1 change

Discussion