Support-Proximity Augmented Diffusion Estimation for Offline Black-Box Optimization

TL;DR

SPADE introduces a novel diffusion-based surrogate model for offline black-box optimization, effectively addressing out-of-distribution issues through support-proximity regularization and calibrated estimation.

Contribution

The paper proposes SPADE, a diffusion model-based framework with new regularization techniques that improve uncertainty quantification and out-of-distribution extrapolation in black-box optimization.

Findings

SPADE achieves state-of-the-art results on Design-Bench tasks.

SPADE outperforms existing methods on LLM data mixture optimization.

Theoretical proof links regularization to Bayesian posterior maximization.

Abstract

Offline black-box optimization aims to discover novel designs with high property scores using only a static dataset, a task fundamentally challenged by the out-of-distribution (OOD) extrapolation problem. Existing approaches typically bifurcate into inverse methods, which struggle with the ill-posed nature of mapping scores to designs, and forward methods, which often lack the distributional expressivity to quantify uncertainty effectively. In this work, we propose SPADE (Support-Proximity Augmented Diffusion Estimation), a novel framework that reimagines forward surrogate modeling through the lens of conditional generative modeling. SPADE models the forward likelihood p(y|x) using a diffusion model, but with two critical enhancements to tailor it for optimization: (1) a Calibrated Diffusion Estimation module that enforces global consistency in statistical moments and pairwise rankings, and (2) a Support-Proximity Regularization mechanism that implicitly internalizes the data manifold constraint p(x) via kNN-based density estimation. Theoretically, we prove that our regularization is first-order equivalent to maximizing a Bayesian posterior with a valid design prior. Empirically, SPADE achieves state-of-the-art performance across Design-Bench tasks and an LLM data mixture optimization benchmark.