Improved Constrained Generation by Bridging Pretrained Generative Models

TL;DR

This paper introduces a framework for constrained generative modeling that fine-tunes pretrained models to generate samples within complex feasible regions, balancing constraint adherence and sample realism.

Contribution

It proposes a novel fine-tuning approach for pretrained generative models to handle complex constraints in structured spatial domains.

Findings

Effective constraint enforcement in complex regions

Maintains high realism in generated samples

Balances constraint satisfaction with sampling quality

Abstract

Constrained generative modeling is fundamental to applications such as robotic control and autonomous driving, where models must respect physical laws and safety-critical constraints. In real-world settings, these constraints rarely take the form of simple linear inequalities, but instead complex feasible regions that resemble road maps or other structured spatial domains. We propose a constrained generation framework that generates samples directly within such feasible regions while preserving realism. Our method fine-tunes a pretrained generative model to enforce constraints while maintaining generative fidelity. Experimentally, our method exhibits characteristics distinct from existing fine-tuning and training-free constrained baselines, revealing a new compromise between constraint satisfaction and sampling quality.