VibeMathedMath problems solved with AI

Petz's conjecture: the scalar curvature of the Kubo-Mori metric increases as a state becomes more mixed

On invertible n×nn\times n density matrices, call ρ1\rho_1 more mixed than ρ2\rho_2 if for each kk the sum of the kk largest eigenvalues of ρ1\rho_1 is at most that of ρ2\rho_2. Petz conjectured that the scalar curvature of the Kubo-Mori (Bogoliubov-Kubo-Mori) metric is monotone under this order: Scal(ρ1)≥Scal(ρ2)\mathrm{Scal}(\rho_1)\ge\mathrm{Scal}(\rho_2) whenever ρ1\rho_1 is more mixed than ρ2\rho_2. In particular the maximum would be at the maximally mixed state.

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI co-developed
Method
Argument
Field
Quantum information geometry; monotone metrics
Posed by
Dénes Petz, Geometry of canonical correlation on the state space of a quantum system, J. Math. Phys. 35 (1994)
Year posed
1994
Years open
32y
Solved
2026-10-09
Model
OpenAI GPT Pro
Vendor
OpenAI
Collaborators
Suvrit Sra
Verification
Unreviewed
Publication
Preprint
Significance
15 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Proves Dittmann's Assertion: hh extends real-analytically to (0,∞)3(0,\infty)^3 and is strictly concave, h(λX+(1−λ)Y)>λh(X)+(1−λ)h(Y)h(\lambda X+(1-\lambda)Y)>\lambda h(X)+(1-\lambda)h(Y) for distinct X,YX,Y and 0<λ<10<\lambda<1. With Dittmann's Theorem 3, used as published, this gives Petz's conjecture. The key step writes h=−1/(8M)h=-1/(8M) with MM homogeneous of degree one and shows t↦1/M(X+tY)t\mapsto1/M(X+tY) is a Stieltjes function.

What the AI did

From the Hexagon provenance statement: "The results in this paper were developed with an AI assistant (OpenAI's GPT Pro). None of the claims were obtained zero-shot: they emerged through sustained back-and-forth discussion, including failed attempts, counterexamples, audits, and repairs." Hexagon labels the text a mix of human-written and AI-generated, with human understanding of all parts.

Verification

Read by this site on 15 November 2026: the Hexagon record and PDF (hexagon:2610.00717v1) and the author's timestamped repository (v1 committed 9 October 2026). The conjecture and the Assertion were compared with Dittmann's own source (arXiv quant-ph/9906009): the paper's formula for hh agrees with Dittmann's symmetrised function to 57 digits on 2,000 random points. An independent check written here found the concavity inequality satisfied on 20,000 random triples of points and weights at 60-digit precision, which supports but does not prove it. The proof itself was not refereed, and Dittmann's Theorem 3 is used as published. No expert reaction found.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion