Action-Inspired Generative Models

TL;DR

This paper presents Action-Inspired Generative Models (AGMs), a novel dual-network framework that improves generative quality by scoring and modulating transport paths without added inference overhead.

Contribution

It introduces a lightweight learned potential that guides the transport process, enhancing existing bridge-matching methods with minimal computational cost.

Findings

Improved generation quality across fidelity and coverage metrics.

The learned potential adds no inference overhead and is a plug-and-play enhancement.

The potential comprises only 1.4% of the primary network's parameters.

Abstract

We introduce Action-Inspired Generative Models (AGMs), a dual-network generative framework motivated by the observation that existing bridge-matching methods assign uniform regression weight to every stochastic transition in the transport landscape, regardless of whether a given bridge sample lies along a structurally coherent trajectory or a degenerate one. We address this by introducing a lightweight learned scalar potential $V_\phi$ that scores bridge samples online and modulates the drift objective via importance weights derived through a stop-gradient barrier -- preventing adversarial feedback between the two networks whilst preserving $V_\phi$'s guiding signal. Crucially, $V_\phi$ comprises only $\sim$1.4% of the primary drift network's parameter count, adds no overhead to the inference graph, and requires no iterative half-bridge fitting or auxiliary stochastic differential equation (SDE) solvers: it is a plug-and-play enhancement to any bridge-matching training loop. At inference, $V_\phi$ is discarded entirely, leaving standard Euler-Maruyama integration of the exponential moving average (EMA) drift. We demonstrate that selectively penalising uninformative transport paths through the learned potential yields consistent improvements in generation quality across fidelity and coverage metrics.