Elliptic Sampling: Optimal Interpolant Schedules for Gaussians and Beyond
arXiv cs.AI · 14 Nov 2026, 04:49 AEST
A new framework for generative modeling based on stochastic interpolants has been introduced, focusing on optimal interpolant schedules and efficient discretization. The framework leads to the development of Zoloflow, a sampling algorithm that achieves minimax optimality for Gaussian targets.
Key facts
- Generative modeling based on stochastic interpolants requires specifying two key choices: an interpolant schedule bridging the initial and target distributions, and an efficient discretization of the resulting probability-flow ODE.
- Our formulation can be solved exactly via conformal mapping, yielding a natural family of variance-preserving elliptic schedules, parametrized by Jacobi elliptic functions.
- For Gaussian targets, Zoloflow achieves minimax optimality among all oblivious score-span algorithms, attaining total-variation error ε using Θ(logκ log(√d/ε)) (exact) score queries.
- For the log-concave case, the elliptic schedule allows sampling with at most ε 2-Wasserstein-error by requiring merely 𝓞(√d polylog(κ, 1/ε)) score evaluations.
Read aloud
A new framework for generative modeling based on stochastic interpolants has been introduced, focusing on optimal interpolant schedules and efficient discretization. The framework leads to the development of Zoloflow, a sampling algorithm that achieves minimax optimality for Gaussian targets. Generative modeling based on stochastic interpolants requires specifying two key choices: an interpolant schedule bridging the initial and target distributions, and an efficient discretization of the resulting probability-flow ODE. Our formulation can be solved exactly via conformal mapping, yielding a natural family of variance-preserving elliptic schedules, parametrized by Jacobi elliptic functions. For Gaussian targets, Zoloflow achieves minimax optimality among all oblivious score-span algorithms, attaining total-variation error ε using Θ logκ log √d/ε exact score queries. For the log-concave case, the elliptic schedule allows sampling with at most ε two-Wasserstein-error by requiring merely 𝓞 √d polylog κ, one/ε score evaluations.
Sources in this cluster
- Elliptic Sampling: Optimal Interpolant Schedules for Gaussians and Beyond
arXiv cs.AI · 14 Nov 2026, 04:49 AEST