A Lagrangian View of Flow Matching

Lipman, Liu

generative modelsoptimal transportpartial differential equations

2609.00198

Builder Relevance

70%

2h ago

Abstract

This paper presents a bottom-up mechanical derivation of Flow Matching grounded in a Lagrangian perspective, emphasizing the conservation of target identity.

Reality Card

Core Claim

The paper establishes a strict invariance condition for optimal single-step generation, leading to a governing quasi-linear advection PDE that explains the straight-line trajectories of Flow Matching.

Method / Result

Solving the governing PDE via the Method of Characteristics yields straight-line trajectories, enabling massive step sizes in generative models.

Limitations

The requirement for distillation to flatten intersecting characteristics may limit the reproducibility of empirical models.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.

← Back to all papers