OctoInformBeta

OpenAI preprint, September 2026 · §1. The variational problem and the main theorem

The zero-temperature Parisi problem: 𝒫(γ)=Φγ(0,0)−12∫01tγ(t)dt\mathcal{P}(\gamma) = \Phi_\gamma(0,0) - \frac12\int_0^1 t\gamma(t)\,dt, with Φγ\Phi_\gamma the value function of a stochastic control problem with terminal reward |x||x|

Minimizing the Parisi functional over zero-temperature order parameters.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

The Sherrington-Kirkpatrick Hamiltonian on configurations σ∈{−1,1}N\sigma \in \{-1,1\}^N is HN(σ)=(2N)−1/2∑i,j=1NgijσiσjH_N(\sigma) = (2N)^{-1/2} \sum_{i,j=1}^N g_{ij} \sigma_i \sigma_j, with independent standard Gaussian couplings gijg_{ij}, so that two configurations with overlap RR have covariance Nξ(R)N\xi(R), where ξ(t)=t2/2\xi(t) = t^2/2. The ground state energy of the Sherrington-Kirkpatrick model is the limit of N−1𝔼max⁡σHN(σ)N^{-1}\,\mathbb{E} \max_\sigma H_N(\sigma) as N→∞N \to \infty. The zero-temperature Parisi formula of Auffinger and Chen identifies this limit with the infimum of the Parisi functional 𝒫(γ)\mathcal{P}(\gamma) over order parameters γ\gamma, the nonnegative, nondecreasing, right-continuous functions on [0,1)[0,1) with finite integral.

Brownian Motion and Bounded Controls

A Brownian system WW on a measurable space Ω\Omega is a probability measure ℙ\mathbb{P} on Ω\Omega, its law, together with a family B=(Bt)t≥0B = (B_t)_{t \ge 0} of real random variables on Ω\Omega that is a standard Brownian motion under ℙ\mathbb{P}.✦ Expectation under ℙ\mathbb{P} is written 𝔼\mathbb{E}.

The increment filtration of WW after a time t≥0t \ge 0 is the filtration (ℱst)s≥0(\mathcal{F}^t_s)_{s \ge 0} on Ω\Omega generated by the increments of BB after tt: ℱst=σ(Bt+r−Bt:0≤r≤s).\mathcal{F}^t_s = \sigma\bigl(B_{t+r} - B_t : 0 \le r \le s\bigr).(1)✦

A bounded control from time tt is a real process α=(αr)r≥0\alpha = (\alpha_r)_{r \ge 0} on Ω\Omega, indexed by the time rr elapsed since tt, that is progressively measurable for (ℱrt)r≥0(\mathcal{F}^t_r)_{r \ge 0} and satisfies |αr(ω)|≤1|\alpha_r(\omega)| \le 1 for every r≥0r \ge 0 and every ω∈Ω\omega \in \Omega.✦

The Definition

Definition 1. Let WW be a Brownian system on Ω\Omega, with law ℙ\mathbb{P} and Brownian motion BB.

An order parameter is a function γ\gamma on the time interval [0,1)[0,1) that is nonnegative, nondecreasing, right-continuous and integrable: 0≤γ(s)≤γ(t)=limr↓tγ(r)(0≤s≤t<1),∫01γ(t)dt<∞.0 \le \gamma(s) \le \gamma(t) = \lim_{r \downarrow t} \gamma(r) \quad (0 \le s \le t < 1), \qquad \int_0^1 \gamma(t)\,dt < \infty.(2)✦ The extension by zero of an order parameter γ\gamma is the function on ℝ\mathbb{R} equal to γ\gamma on [0,1)[0,1) and to 00 elsewhere; integrals of γ\gamma over intervals of ℝ\mathbb{R} are those of this function.

The value function of an order parameter γ\gamma is the function Φγ\Phi_\gamma on [0,1]×ℝ[0,1] \times \mathbb{R} whose value at a time tt and a position xx is the supremum, over bounded controls α\alpha from time tt, of the expected control payoff: Φγ(t,x)=supα𝔼[|x+B1−Bt+∫t1γ(s)αs−tds|−12∫t1γ(s)αs−t2ds].\Phi_\gamma(t,x) = \sup_\alpha \mathbb{E}\Bigl[\Bigl|x + B_1 - B_t + \int_t^1 \gamma(s)\,\alpha_{s-t}\,ds\Bigr| - \frac12 \int_t^1 \gamma(s)\,\alpha_{s-t}^2\,ds\Bigr].(3)✦

The Parisi functional 𝒫\mathcal{P} is the function on order parameters that assigns to each order parameter γ\gamma the number 𝒫(γ)=Φγ(0,0)−12∫01tγ(t)dt.\mathcal{P}(\gamma) = \Phi_\gamma(0,0) - \frac12 \int_0^1 t\,\gamma(t)\,dt.(4)✦

An order parameter γ\gamma is a Parisi minimizer, a solution of the zero-temperature Parisi problem, when 𝒫(γ)≤𝒫(η)for every order parameter η.\mathcal{P}(\gamma) \le \mathcal{P}(\eta) \quad \text{for every order parameter } \eta.(5)✦

What Is Checked

The definition is stated over a Brownian system WW on Ω\Omega, the increment filtrations (1) of its Brownian motion and the bounded controls progressively measurable for them, the control payoff inside (3), and the extension by zero of an order parameter from the time interval [0,1)[0,1) to the real line.