Black-Box Optimization From Small Offline Datasets via Meta Learning with Synthetic Tasks

Azza Fadhel, The Hung Tran, Trong Nghia Hoang, Jana Doppa
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:3862-3870, 2026.

Abstract

We consider the problem of offline black-box optimization, where the goal is to discover optimal designs (e.g., molecules or materials) from past experimental data. A key challenge in this setting is data scarcity: in many scientific applications, only small or poor-quality datasets are available, which severely limits the effectiveness of existing algorithms. Prior work has theoretically and empirically shown that performance of offline optimization algorithms depends on how well the surrogate model captures the optimization bias (i.e., ability to rank input designs correctly), which is challenging to accomplish with limited experimental data. This paper proposes {\em Surrogate Learning with Optimization Bias via Synthetic Task Generation} (\textsc{OptBias}), a meta-learning framework that directly tackles data scarcity. OptBias learns a reusable optimization bias by training on synthetic tasks generated from a Gaussian process, and then fine-tunes the surrogate model on the small data for the target task. Across diverse continuous and discrete offline optimization benchmarks, OptBias consistently outperforms state-of-the-art baselines in small data regimes. These results highlight OptBias as a robust and practical solution for offline optimization in realistic small data settings.

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-fadhel26a, title = { Black-Box Optimization From Small Offline Datasets via Meta Learning with Synthetic Tasks }, author = {Fadhel, Azza and Tran, The Hung and Hoang, Trong Nghia and Doppa, Jana}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {3862--3870}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/fadhel26a/fadhel26a.pdf}, url = {https://proceedings.mlr.press/v300/fadhel26a.html}, abstract = { We consider the problem of offline black-box optimization, where the goal is to discover optimal designs (e.g., molecules or materials) from past experimental data. A key challenge in this setting is data scarcity: in many scientific applications, only small or poor-quality datasets are available, which severely limits the effectiveness of existing algorithms. Prior work has theoretically and empirically shown that performance of offline optimization algorithms depends on how well the surrogate model captures the optimization bias (i.e., ability to rank input designs correctly), which is challenging to accomplish with limited experimental data. This paper proposes {\em Surrogate Learning with Optimization Bias via Synthetic Task Generation} (\textsc{OptBias}), a meta-learning framework that directly tackles data scarcity. OptBias learns a reusable optimization bias by training on synthetic tasks generated from a Gaussian process, and then fine-tunes the surrogate model on the small data for the target task. Across diverse continuous and discrete offline optimization benchmarks, OptBias consistently outperforms state-of-the-art baselines in small data regimes. These results highlight OptBias as a robust and practical solution for offline optimization in realistic small data settings. } }
Endnote
%0 Conference Paper %T Black-Box Optimization From Small Offline Datasets via Meta Learning with Synthetic Tasks %A Azza Fadhel %A The Hung Tran %A Trong Nghia Hoang %A Jana Doppa %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-fadhel26a %I PMLR %P 3862--3870 %U https://proceedings.mlr.press/v300/fadhel26a.html %V 300 %X We consider the problem of offline black-box optimization, where the goal is to discover optimal designs (e.g., molecules or materials) from past experimental data. A key challenge in this setting is data scarcity: in many scientific applications, only small or poor-quality datasets are available, which severely limits the effectiveness of existing algorithms. Prior work has theoretically and empirically shown that performance of offline optimization algorithms depends on how well the surrogate model captures the optimization bias (i.e., ability to rank input designs correctly), which is challenging to accomplish with limited experimental data. This paper proposes {\em Surrogate Learning with Optimization Bias via Synthetic Task Generation} (\textsc{OptBias}), a meta-learning framework that directly tackles data scarcity. OptBias learns a reusable optimization bias by training on synthetic tasks generated from a Gaussian process, and then fine-tunes the surrogate model on the small data for the target task. Across diverse continuous and discrete offline optimization benchmarks, OptBias consistently outperforms state-of-the-art baselines in small data regimes. These results highlight OptBias as a robust and practical solution for offline optimization in realistic small data settings.
APA
Fadhel, A., Tran, T.H., Hoang, T.N. & Doppa, J.. (2026). Black-Box Optimization From Small Offline Datasets via Meta Learning with Synthetic Tasks . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:3862-3870 Available from https://proceedings.mlr.press/v300/fadhel26a.html.

Related Material