Efficient Personalization of Generative Models via Optimal Experimental Design
This addresses the costly and time-consuming nature of obtaining human feedback for aligning generative models with end-user needs, representing an incremental improvement in data efficiency.
The paper tackles the problem of efficiently personalizing generative models to user preferences by developing an optimal experimental design approach that selects the most informative preference queries, requiring fewer queries than random selection for text-to-image model personalization.
Preference learning from human feedback has the ability to align generative models with the needs of end-users. Human feedback is costly and time-consuming to obtain, which creates demand for data-efficient query selection methods. This work presents a novel approach that leverages optimal experimental design to ask humans the most informative preference queries, from which we can elucidate the latent reward function modeling user preferences efficiently. We formulate the problem of preference query selection as the one that maximizes the information about the underlying latent preference model. We show that this problem has a convex optimization formulation, and introduce a statistically and computationally efficient algorithm ED-PBRL that is supported by theoretical guarantees and can efficiently construct structured queries such as images or text. We empirically present the proposed framework by personalizing a text-to-image generative model to user-specific styles, showing that it requires less preference queries compared to random query selection.