pith. sign in

arxiv: 2606.23383 · v1 · pith:3UQ5KLMUnew · submitted 2026-06-22 · 🧮 math.PR · math.FA· stat.ML

Near-Lipschitz stability of the Kim--Milman flow map

Sinho Chewi , Katharina Eichinger , Aram-Alexandre Pooladian This is my paper

Pith reviewed 2026-06-26 06:58 UTC · model grok-4.3

classification 🧮 math.PR math.FAstat.ML
keywords Kim-Milman flow mapstability2-Wasserstein distancerelative entropyoptimal transportprobability measuresflow map
0
0 comments X

The pith

The Kim-Milman flow map is stable in relative entropy and Lipschitz stable in 2-Wasserstein distance up to a logarithmic factor when one target measure is regular.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proves that the Kim-Milman flow map between probability measures varies continuously when the target measure is perturbed, as long as one of the measures meets a regularity condition. Stability is shown both in relative entropy and, more strongly, as near-Lipschitz continuity in the 2-Wasserstein distance. A separate existence result establishes the map for every target measure that has finite second moment. These properties matter for any setting that relies on the map to move mass from one distribution to another, since they guarantee that small changes or approximations in the target do not produce large changes in the resulting map.

Core claim

The Kim--Milman flow map enjoys favorable stability properties with respect to variations in the target measure, provided that one of the target measures is sufficiently regular. The results include stability in relative entropy, and Lipschitz stability in the 2-Wasserstein distance up to a logarithmic factor. A general existence theorem holds for these maps for any target measure with finite second moment.

What carries the argument

The Kim--Milman flow map, the continuous-time path that deforms one probability measure into another while preserving the optimal-transport structure at each step.

If this is right

  • The flow map remains close in relative entropy when the target is changed slightly under the regularity assumption.
  • The map is Lipschitz continuous in the 2-Wasserstein metric up to a logarithmic factor.
  • The map exists for every target measure possessing a finite second moment.
  • Approximations of the target by regular measures produce correspondingly small changes in the flow map.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Numerical schemes that approximate the target measure by a regular one can inherit error bounds from the stability result.
  • The result may support robustness analyses when the target is replaced by an empirical measure drawn from data.
  • Similar near-Lipschitz statements could be tested for other continuous interpolations between optimal transport maps.

Load-bearing premise

One of the two target probability measures must satisfy a sufficient regularity condition.

What would settle it

An explicit pair of irregular target measures whose Kim-Milman flow maps differ by more than a logarithmic factor in 2-Wasserstein distance after a small perturbation of the target.

read the original abstract

We prove that the Kim--Milman flow map enjoys favorable stability properties with respect to variations in the target measure, provided that one of the target measures is sufficiently regular. Our results include stability in relative entropy, and more notably, Lipschitz stability in the $2$-Wasserstein distance up to a logarithmic factor. We complement our results with a general existence theorem for these maps for any target measure with finite second moment.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The manuscript proves that the Kim--Milman flow map enjoys stability with respect to perturbations of the target measure when at least one target satisfies a regularity condition. The results include stability in relative entropy and near-Lipschitz stability (up to a logarithmic factor) in the 2-Wasserstein distance. A separate existence result is given for any target measure possessing finite second moment.

Significance. If the claims hold, the near-Lipschitz Wasserstein stability would be a useful quantitative robustness result for these flow maps in the theory of optimal transport and related gradient flows.

minor comments (2)
  1. [Abstract] Abstract and §1: the phrase 'sufficiently regular' is used without an explicit statement of the condition (e.g., bounded density, log-concavity, or a specific Sobolev norm). This should be stated precisely at the first appearance.
  2. [Introduction] The existence theorem is stated for targets with finite second moment, but the stability theorems require an additional regularity assumption on one measure; the manuscript should clarify whether the existence result is used as a stepping stone or is independent.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of our work and for recommending minor revision. We appreciate the acknowledgment that the near-Lipschitz Wasserstein stability result would be a useful quantitative robustness statement in optimal transport.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The provided abstract and context describe a theoretical stability result for the Kim-Milman flow map under a regularity assumption on one target measure, plus an existence theorem for measures with finite second moment. No derivation steps, equations, self-citations, fitted parameters, or ansatzes are exhibited that reduce any claimed prediction or result to its own inputs by construction. The work is self-contained against external benchmarks as a pure existence/stability proof in optimal transport, with no load-bearing reductions detectable from the given material.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no free parameters, axioms, or invented entities can be identified from the given text.

pith-pipeline@v0.9.1-grok · 5597 in / 1002 out tokens · 26850 ms · 2026-06-26T06:58:17.003743+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

16 extracted references · 2 linked inside Pith

  1. [1]

    Stability bounds for smooth optimal transport maps and their statistical implications.arXiv preprint 2502.12326,

    Sivaraman Balakrishnan and Tudor Manole. Stability bounds for smooth optimal transport maps and their statistical implications.arXiv preprint 2502.12326,

    arXiv

  2. [2]

    Statistical estimation of Monge transport maps via Brenier potentials.arXiv preprint 2604.22366,

    Elsa Cazelles, Edouard Pauwels, and L´ eo Portales. Statistical estimation of Monge transport maps via Brenier potentials.arXiv preprint 2604.22366,

    Pith/arXiv arXiv

  3. [3]

    Propagation of weak log-concavity along generalised heat flows via Hamilton–Jacobi equations.arXiv preprint 2508.07931, August

    Louis-Pierre Chaintron, Giovanni Conforti, and Katharina Eichinger. Propagation of weak log-concavity along generalised heat flows via Hamilton–Jacobi equations.arXiv preprint 2508.07931, August

    arXiv

  4. [4]

    Sinho Chewi, Aram-Alexandre Pooladian, and Matthew S. Zhang. Stability of the Kim– Milman flow map.arXiv preprint 2511.01154,

    Pith/arXiv arXiv

  5. [5]

    A coupling approach to Lipschitz transport maps.arXiv preprint 2502.01353,

    Giovanni Conforti and Katharina Eichinger. A coupling approach to Lipschitz transport maps.arXiv preprint 2502.01353,

    arXiv

  6. [6]

    Tight stability bounds for entropic Brenier maps.International Mathematics Research Notices, 2025(7):rnaf078,

    17 Vincent Divol, Jonathan Niles-Weed, and Aram-Alexandre Pooladian. Tight stability bounds for entropic Brenier maps.International Mathematics Research Notices, 2025(7):rnaf078,

    2025

  7. [7]

    An entropy approach to the time reversal of diffusion processes

    Hans F¨ ollmer. An entropy approach to the time reversal of diffusion processes. InStochastic differential systems (Marseille-Luminy, 1984), volume 69 ofLect. Notes Control Inf. Sci., pages 156–163. Springer, Berlin,

    1984

  8. [8]

    A generalization of Caffarelli’s contraction theorem to nearly spherical manifolds.arXiv preprint 2512.01496,

    Yuxin Ge and Jordan Serres. A generalization of Caffarelli’s contraction theorem to nearly spherical manifolds.arXiv preprint 2512.01496,

    arXiv

  9. [9]

    Global regularity estimates for optimal transport via entropic regularisation.arXiv preprint 2501.11382,

    Nathael Gozlan and Maxime Sylvestre. Global regularity estimates for optimal transport via entropic regularisation.arXiv preprint 2501.11382,

    arXiv

  10. [10]

    Ksenia A. Khudiakova, Jan Maas, and Francesco Pedrotti.L∞-optimal transport of anisotropic log-concave measures and exponential convergence in Fisher’s infinitesimal model.The Annals of Applied Probability, 35(3):1913–1940,

    1913

  11. [11]

    Stability of optimal transport maps on Riemannian manifolds.arXiv preprint 2504.05412,

    Jun Kitagawa, Cyril Letrouit, and Quentin M´ erigot. Stability of optimal transport maps on Riemannian manifolds.arXiv preprint 2504.05412,

    arXiv

  12. [12]

    Quantitative stability of optimal transport.Notes du cours Peccot, 2025,

    Cyril Letrouit. Quantitative stability of optimal transport.Notes du cours Peccot, 2025,

    2025

  13. [13]

    Gluing methods for quantitative stability of optimal transport maps.arXiv preprint 2411.04908,

    18 Cyril Letrouit and Quentin M´ erigot. Gluing methods for quantitative stability of optimal transport maps.arXiv preprint 2411.04908,

    arXiv

  14. [14]

    Lipschitz changes of variables via heat flow.arXiv preprint 2201.03403,

    Joe Neeman. Lipschitz changes of variables via heat flow.arXiv preprint 2201.03403,

    arXiv

  15. [15]

    Entropic estimation of optimal transport maps.arXiv preprint 2109.12004,

    Aram-Alexandre Pooladian and Jonathan Niles-Weed. Entropic estimation of optimal transport maps.arXiv preprint 2109.12004,

    arXiv

  16. [16]

    µ(dy). Writing x = R + s for −R1/3 ⩽s⩽R 1/3 and y = R/(1 + t) + u for u∈ [0, R−1], we have the bound |y−x| 2 − |x−R| 2 = tR 1 +t −u tR 1 +t −u+ 2s ⩾ tR 1 +t −R −1 tR 1 +t −R −1 −2R 1/3 , 26 which is strictly positive forRlarge enough. As a result, α(x) pRβ(x) =o(1), with exponentially fast decay. These two terms imply that νt(x) pRβ(x) = 1 +o(1). Note tha...

    2000