A General Framework for Inference-time Scaling and Steering of Diffusion Models
This work addresses the problem of controllability and efficiency in diffusion models for researchers and practitioners, offering a novel inference-time method that avoids expensive training and mode collapse.
The paper tackles the challenge of generating samples with user-specified properties in diffusion models by introducing Feynman-Kac (FK) steering, an inference-time framework that uses reward functions to steer diffusion processes without fine-tuning. The results show that FK steering with a 0.8B parameter model outperforms a 2.6B fine-tuned model in prompt fidelity for text-to-image tasks and generates lower perplexity, more linguistically acceptable outputs for text diffusion models.
Diffusion models produce impressive results in modalities ranging from images and video to protein design and text. However, generating samples with user-specified properties remains a challenge. Recent research proposes fine-tuning models to maximize rewards that capture desired properties, but these methods require expensive training and are prone to mode collapse. In this work, we present Feynman-Kac (FK) steering, an inference-time framework for steering diffusion models with reward functions. FK steering works by sampling a system of multiple interacting diffusion processes, called particles, and resampling particles at intermediate steps based on scores computed using functions called potentials. Potentials are defined using rewards for intermediate states and are selected such that a high value indicates that the particle will yield a high-reward sample. We explore various choices of potentials, intermediate rewards, and samplers. We evaluate FK steering on text-to-image and text diffusion models. For steering text-to-image models with a human preference reward, we find that FK steering a 0.8B parameter model outperforms a 2.6B parameter fine-tuned model on prompt fidelity, with faster sampling and no training. For steering text diffusion models with rewards for text quality and specific text attributes, we find that FK steering generates lower perplexity, more linguistically acceptable outputs and enables gradient-free control of attributes like toxicity. Our results demonstrate that inference-time scaling and steering of diffusion models - even with off-the-shelf rewards - can provide significant sample quality gains and controllability benefits. Code is available at https://github.com/zacharyhorvitz/Fk-Diffusion-Steering .