# Sample-Efficient Optimisation over the Outputs of Generative Models

Source: [Sample-Efficient Optimisation over the Outputs of Generative Models](<https://www.sota2.com/research/paper/68fa7bb41e209131cea8a06f>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-15T05:00:12.430Z

Authors: Samuel Willis, Paul Duckworth, Jack Simons, Aleksandra Kalisz, Krisztina Sinkovics, Noam Ghenassia, Shikha Surana, Henry T. Oldroyd, Alexandru I. Stere, Dragos D Margineantu, Carl Henrik Ek, Henry Moss, Erik Bodin

Published: 2025-09-28T10:50:06.000Z

## Abstract

Modern generative AI models, such as diffusion and flow matching models, can sample from rich data distributions. However, many applications, especially in science and engineering, require more than drawing samples from the model distribution: they require searching within this distribution for samples that optimise task-specific criteria. In this work, we propose O3 (Optimisation Over the Outputs of Generative Models), a method for sample-efficient black-box optimisation over continuous-variable diffusion and flow-matching models. O3 is built around surrogate latent spaces: low-dimensional Euclidean embeddings that can be extracted from a generative model without additional training. The resulting representations have controllable dimensionality and support the direct application of standard optimisation algorithms. We show, on image and protein design tasks, that surrogate-space optimisation finds substantially higher-scoring samples than standard sampling or optimisation in the original latent space. Our method is model- and optimiser-agnostic, incurs negligible additional cost over standard generation, and requires no retraining or fine-tuning of the generative model.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Text-to-Image Generation on Stable Diffusion A green colored rabbit. 1.5](<https://www.sota2.com/research/sota/text-to-image-generation-on-stable-diffusion-a-green-colored-rabbit-1-5.md>): Text-to-Image Generation; Stable Diffusion A green colored rabbit. 1.5; Reward: 1.75.

- [Text-to-Image Generation on Stable Diffusion Two roses in a vase 1.5](<https://www.sota2.com/research/sota/text-to-image-generation-on-stable-diffusion-two-roses-in-a-vase-1-5.md>): Text-to-Image Generation; Stable Diffusion Two roses in a vase 1.5; Reward: 1.55.

- [Text-to-Image Generation on Stable Diffusion "Two dogs in the park." 1.5](<https://www.sota2.com/research/sota/text-to-image-generation-on-stable-diffusion-two-dogs-in-the-park-1-5.md>): Text-to-Image Generation; Stable Diffusion "Two dogs in the park." 1.5; Reward: 0.84.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2509.23800>)

[PDF](<https://arxiv.org/pdf/2509.23800.pdf>)
