Conflict-Aware Additive Guidance for Flow Models under Compositional Rewards

TL;DR

This paper introduces Conflict-Aware Additive Guidance ($g^\text{car}$), a lightweight method that dynamically detects and resolves gradient conflicts to improve controlled generation in flow models with multiple constraints.

Contribution

It proposes a novel, learnable guidance method that actively rectifies off-manifold drift caused by gradient conflicts in compositional constraints.

Findings

$g^\text{car}$ surpasses baselines in generation fidelity.

It effectively rectifies off-manifold drift across diverse domains.

Code is available at https://github.com/yuxuehui/CAR-guidance.

Abstract

Inference-time guided sampling steers state-of-the-art diffusion and flow models without fine-tuning by interpreting the generation process as a controllable trajectory. This provides a simple and flexible way to inject external constraints (e.g., cost functions or pre-trained verifiers) for controlled generation. However, existing methods often fail when composing multiple constraints simultaneously, which leads to deviations from the true data manifold. In this work, we identify root causes of this off-manifold drift and find that the approximation error scales severely with gradient misalignment. Building on these findings, we propose Conflict-Aware Additive Guidance ($g^\text{car}$), a lightweight and learnable method, which actively rectifies off-manifold drift by dynamically detecting and resolving gradient conflicts. We validate $g^\text{car}$ across diverse domains, ranging from synthetic datasets and image editing to generative decision-making for planning and control. Our results demonstrate that $g^\text{car}$ effectively rectifies off-manifold drift, surpassing baselines in generation fidelity while using light compute. Code is available at https://github.com/yuxuehui/CAR-guidance.