VibeMathedMath problems solved with AI

Petty's polar projection problem

For a convex body K⊂RnK\subset\mathbb{R}^n the projection body ΠK\Pi K has support function hΠK(u)=voln−1(K∣u⊥)h_{\Pi K}(u)=\mathrm{vol}_{n-1}(K|u^\perp), and its polar Π∗K=(ΠK)∘\Pi^*K=(\Pi K)^\circ is the polar projection body. For an origin-centred ellipsoid KK, Π∗K\Pi^*K is a dilate of KK. Petty's polar projection problem, which Horvath and Langi call a famous problem of convex geometry and state as a conjecture, asks for the converse: if an origin-symmetric convex body K⊂RnK\subset\mathbb{R}^n, n≥3n\ge3, satisfies Π∗K=λK\Pi^*K=\lambda K for some λ>0\lambda>0, must KK be an ellipsoid? In the plane the answer is no (Radon curves, for example the regular hexagon). Horvath and Langi (2022) showed the problem is equivalent to their translative constant volume conjecture: a centrally symmetric body for which the volume of the convex hull of two touching translates does not depend on the pair is an ellipsoid.

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI co-developed
Method
Argument
Field
Convex geometry; projection bodies
Posed by
C. M. Petty, Isoperimetric problems, Proc. Conf. Convexity and Combinatorial Geometry, Univ. of Oklahoma (1971)
Year posed
1971
Years open
55y
Solved
2026-11-12
Model
GPT-6 Astra (ChatGPT)
Vendor
OpenAI
Collaborators
Sihan Chen
Verification
Unreviewed
Publication
Preprint
Significance
26 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1: for n≥3n\ge3 and a convex body K⊂RnK\subset\mathbb{R}^n, with TK=(ΠK)∘TK=(\Pi K)^\circ, the following are equivalent: KK is an origin-centred ellipsoid; TK=aKTK=aK for some a>0a>0; T2K=bKT^2K=bK for some b>0b>0. No symmetry is assumed, since TKTK is always origin-symmetric. The proof has three steps: a flux identity ∫hL−1 dSL=∫AL dSK\int h_L^{-1}\,dS_L=\int A_L\,dS_K for L=TKL=TK, where AC(u)A_C(u) is the volume of the central section C∩u⊥C\cap u^\perp (Lemma 2.1); the comparison ∫AK dSL≤∫AL dSK\int A_K\,dS_L\le\int A_L\,dS_K for symmetric KK (Proposition 3.1), whose equality case gives, for every direction, a contractive projection onto a hyperplane along it, so that KK is an ellipsoid by a theorem of James; and the remark that T2K=bKT^2K=bK forces equality. Corollary 5.1 deduces the translative constant volume conjecture of Horvath and Langi. Dimension two is excluded, as it must be.

What the AI did

From the paper: OpenAI's ChatGPT was used to explore proof strategies, develop and check mathematical arguments, and draft and revise the exposition. Hexagon's provenance record names the tool as ChatGPT, GPT 6 Astra, labels the work 'Primarily AI-generated text' with human understanding of 'some parts', and carries the author's statement: the entire derivation was verified and re-derived by the author with no issues found; the author came to the problem while adapting ideas from a manuscript on fixed points of the centre of mass, whose line of reasoning is in the appendix; and 'the final key derivation and ideas were completed by AI'.

Verification

No independent mathematician has checked this yet. Checked here on 19 November 2026: the Hexagon record and PDF of 2611.00401v1 (Sihan Chen, 5 pages, 12 November 2026) were read in full, including the provenance label and the Use of AI section. The problem was read in Horvath and Langi (arXiv 2012.08955), who state it as Conjecture 2 with Petty's 1971 paper as source and prove the equivalence used in Corollary 5.1; Petty's proceedings paper itself was not opened. Each step of the proof was followed and no gap was found; that is this site's reading, not an expert review. Two identities were checked by hand on the cube [−1,1]3[-1,1]^3, where TKTK is the octahedron with vertices ±ei/4\pm e_i/4: both sides of Lemma 2.1 equal 33, and Proposition 3.1 reads 94≤3\frac94\le3, with pointwise equality AK(u)=1/hL(u)A_K(u)=1/h_L(u) at the coordinate directions. The final step rests on James's 1947 characterization of inner product spaces, which was taken as cited. No formalization. Hexagon is moderated but not peer reviewed. The author deposited five primarily AI-generated papers there in November 2026, which is a reason for a specialist to read this one before the status changes.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion