Science Explained/Brief

Two-stage method improves multi-objective optimisation under tight budgets

A new preprint describes a 'converge-then-diversify' approach for multi-objective Bayesian optimisation, reporting that it outperformed existing methods in most of 446 comparisons. The work is not yet peer-reviewed.

BriefPublished 15 September 20261 min read1 linked source · 3 checked facts

Multi-objective Bayesian optimisation (MOBO) is a method for optimising expensive black-box functions with multiple objectives. The goal is to approximate the Pareto front—a set of solutions representing the best trade-offs—with both good convergence (closeness to the front) and good diversity (spread across it). Existing methods try to achieve both at once, but under very tight evaluation budgets, there may not be enough solutions to cover the entire front. The new approach, called converge-then-diversify (CTD), splits the search into two stages: first, it drives toward a single point on the front; second, it spreads solutions across the front. The paper presents two instantiations using widely adopted acquisition functions. In experiments across 446 pairwise comparisons, CTD statistically outperformed state-of-the-art methods in 72.9% of cases, performed equivalently in 21.1%, and was worse in only 6.1%. The advantage was particularly evident with very tight budgets or high-dimensional problems.

What the reporting says

Each line below is stated by the page named above it, and only by that page.

Stated by arXiv

Multi-objective Bayesian optimisation (MOBO) is a sample-efficient approach for optimising expensive black-box functions with multiple objectives.
The paper proposes a converge-then-diversify (CTD) approach that decouples convergence and diversity into two stages.
Experimental results show that, across all 446 pairwise comparisons, CTD statistically outperforms state-of-the-art methods in 72.9% of the cases, performs equivalently in 21.1%, and is statistically worse in only 6.1%.

Our view

This two-stage approach may be particularly useful when evaluation budgets are very limited, but the lack of peer review means the results should be treated as preliminary.

What the reporting says: Across all 446 pairwise comparisons, CTD statistically outperforms state-of-the-art methods in 72.9% of cases, performs equivalently in 21.1%, and is statistically worse in only 6.1%.

Sources

arXivText stored 15 September 2026 · checked 15 September 2026

Found through monitored feeds: arXiv. A feed address is how the desk finds a story, not the evidence for it — the sources listed above are the pages the facts were taken from.

How this story was checked. Written from the 1 page listed above, stored 15 September 2026; claims checked against that stored text on 15 September 2026.

What that means

3 of 3 reported statements were confirmed against the page that carries them; the rest were removed rather than published.
Figures in the text were required to appear in the stored source text: yes. Identifiers: yes.
The check reads stored text only: no claim rests on a fresh look that did not happen.
Where the reporting was silent, the text says so instead of filling the gap.

Two-stage method improves multi-objective optimisation under tight budgets

What the reporting says

Our view

Sources

More from Science Explained

AI agent harness claims top rank across four life science tasks in preprint

Vendor-Native Coding Harnesses Show No Clear Average Advantage in Paired Test

Preprint tests LLM judge for comparing clinical timelines to case reports