VibeMathedMath problems solved with AI

The permanence conjecture for weakly reversible mass-action systems with fixed rates

For a weakly reversible reaction network with mass-action kinetics x˙=∑y→y′ky→y′xy(y′−y)\dot x=\sum_{y\to y'}k_{y\to y'}x^y(y'-y) and fixed positive rates, let SS be the span of the reaction vectors y′−yy'-y and P=(c+S)∩R>0dP=(c+S)\cap\mathbb R^d_{>0} a positive stoichiometric class, which may be unbounded. The system is permanent on PP if every solution starting in PP is global and positive and one compact K⊂PK\subset P eventually contains each of them, the entry time but not KK depending on the initial point. This is stronger than boundedness plus persistence of each trajectory. Known before: permanence for strongly endotactic networks, including weakly reversible ones with one linkage class (Gopalkrishnan-Miller-Shiu; Boros-Hofbauer), and for two species (Craciun-Nazarov-Pantea). Is every weakly reversible mass-action system with fixed positive rates permanent on every positive stoichiometric class?

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI-discovered
Method
Argument
Field
Chemical reaction network theory: mass-action kinetics
Posed by
Permanence conjecture of chemical reaction network theory; the manuscript names it without an originating source and takes the classwise definition from Boros and Hofbauer (2020, Definition 4.1)
Year posed
—
Years open
—
Solved
2026-10-05
Model
Unreleased internal OpenAI model
Vendor
OpenAI
Collaborators
—
Verification
Unreviewed
Publication
Announced
Collection
OpenAI math release (October 2026), version adc7f12
Significance
32 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1: for a finite weakly reversible network with fixed positive rates and any positive stoichiometric class PP, there is a nonempty compact convex forward-invariant KP⊂PK_P\subset P that every solution from PP enters in finite time; hence one εP\varepsilon_P bounds every trajectory in PP between εP\varepsilon_P and εP−1\varepsilon_P^{-1} after its entry time, also on unbounded classes. The proof adds strict increase off the trapping set and a passage between scales to the companion's construction. Uniformity holds within one class and one choice of rates; entry times may depend on the initial point, and time-varying rates are not treated.

What the AI did

The OpenAI math release (github.com/openai/math, commit adc7f12) states that its results were produced by an unreleased internal OpenAI model under one fixed procedure, averaging about three hours of ChatGPT Pro thinking compute per result, across roughly 4,000 posed problems; outputs were then grouped into families and filtered for significance. This result is not among the README's stated exceptions (the Riemann zeta zero-free region work and the Hodge conjecture for CM abelian varieties). The manuscript is credited to OpenAI alone and names no human author. The manuscript credits its affine approximation lemma and trapping construction to the companion Boundedness and persistence manuscript of September 25, also produced by the model; the README notes that some outputs build on earlier model results. The README also cautions that unformalized results could have issues.

Verification

No independent mathematician has checked this yet. Checked here: the abstract, introduction and Theorem 1.1 of the TeX source; the proof was not refereed. The family's only Lean challenge (MassAction) formalizes the companion boundedness-and-persistence theorem, whose bound depends on the initial point, so this classwise permanence statement has no formal proof at the pinned commit. The argument reuses the companion manuscript's construction, which is reproduced in the paper.

Sources

Changelog1 change

Discussion