Generative Refinement for Low-Budget Black-Box Optimization

2026-07-01

AI Summary

This work addresses black-box optimization challenges under low evaluation budgets, unreliable rewards, or when the optimum lies in narrow or disconnected regions—scenarios where conventional methods fail due to their reliance on extensive function evaluations. The authors propose SPARROW, an algorithm that achieves full decoupling between generative priors and reward signals for the first time. SPARROW leverages any pre-trained generative model—trained on reward-free data—as a fixed, structured proposal operator and guides the search using rankings of evaluated candidates. This approach drastically reduces dependence on function evaluations, enabling efficient optimization of complex geometric objectives even under extremely limited budgets. The method provides asymptotic convergence guarantees over the generator’s support and demonstrates superior robustness in environments with noisy or unreliable rewards.

Abstract

Black-box optimization is a fundamental science and engineering tool that makes it possible to optimize objectives without gradient information. Unfortunately, as it often requires many function evaluations, it can be challenging when each one is costly. This is especially true when the evaluation function is noisy or failure-prone, and when high-performing solutions are confined to thin, curved, or disconnected regions of the search space. Existing methods leveraging generative models to navigate these subspaces are built to sample from reward-aligned distributions. As a result, they require a large number of evaluations to align their sampler effectively, making them impractical in low-budget settings. We propose SPARROW, an algorithm that completely decouples the generative prior from the reward signal. SPARROW can use any sampler with a known corruption process and trained on unevaluated data, as a fixed, structured proposal operator. Optimization proceeds by rank-based guidance over an archive of evaluated candidates. SPARROW can navigate complex geometries, handle unreliable reward signals, and perform effective optimization under very low evaluation budgets. We provide asymptotic convergence guarantees over the sampler support and demonstrate strong empirical performance on problems with unreliable rewards and geometrically complex landscapes.

Problem

• black-box optimization
• low-budget
• generative models
• noisy evaluation
• complex search landscapes

Innovation

• black-box optimization
• generative models
• low-budget optimization
• reward-agnostic sampling
• rank-based guidance