Statistical attribute alignment for black-box generative AI via output post-processing

2026-09-25

AI Summary

This study addresses the misalignment between the attribute distribution of black-box generative AI outputs and user-specified targets by proposing a post-processing algorithm based on query strategy optimization. By integrating statistical alignment theory with optimization techniques, the method formulates an optimal strategy that minimizes the expected number of queries, thereby overcoming the limitations of conventional prompt-based interventions and ensuring that the joint attribute distribution of generated outputs precisely matches the target distribution. Experimental evaluations on text-to-image synthesis and data generation tasks demonstrate that the proposed algorithm significantly enhances attribute alignment, consistently outperforming existing baseline methods.

Abstract

Generative AI systems are increasingly used, but aligning their outputs with user requirements poses a continuing challenge. Here, we aim to ensure that the distribution of an attribute of an AI-generated output aligns with a user-specified target. This is motivated by examples such as fairness, where we want to ensure that a protected attribute (e.g., gender, race, or age categories) follows a desired distribution, and synthetic data generation, where we want the generated data to be representative of a target distribution. We study the practically important black-box access setting, where a user can repeatedly query a generative AI model. The goal is to return $m\ge 1$ outputs whose joint attribute distribution is as close as possible to this target. For both exact and approximate alignment, we develop algorithms that minimize the expected number of queries to the generator, and we further demonstrate their optimality as the number of requested outputs $m \rightarrow \infty$. Experiments on text-to-image generation and geocoded persona generation tasks show that our post-processing algorithms improve statistical attribute alignment, complementing prompting-based interventions.

Problem

• generative AI
• black-box access
• attribute alignment
• distribution matching
• output post-processing

Innovation

• Black-box generative AI
• Statistical attribute alignment
• Output post-processing
• Query optimization
• Fairness