aiqbench analysis
The paper applies meta-learning to experiment-driven black-box optimization: a policy is trained on synthetic practice functions built from sparse records of past campaigns, then run on new tasks with frozen parameters. Reported gains are "lowest median final-budget value" against the paper's own baselines, on four of five synthetic families and both engineering benchmarks (BOPTEST, SU2). Readers should check how those baselines were tuned and whether the synthetic-to-real transfer holds before treating this as a general improvement over Bayesian optimization.
CAP: Learning to Optimize from Sparse Observations of Related Tasks
arXiv:2610.41173v1 Announce Type: new Abstract: In experiment-driven black-box optimization, each evaluation can consume materials, equipment time, and other resources, limiting the number of experiments available for a new task. Records from completed related optimization campaigns could help, but learning a sequential query policy remains challenging when task-specific experience consists of only a few sparse histories. We propose CAP (Candidate Acquisition Policy), which models shared structure and task-specific variation in fixed historical records to sample interactive practice functions. On these functions, CAP learns a feedback-conditioned query policy aimed at final optimization performance under a fixed evaluation budget. At deployment, its parameters remain fixed while each new observation informs the next query. Across five synthetic task families spanning 6 and 20 dimensions and two engineering simulation benchmarks, BOPTEST and SU2, CAP attains the lowest median final-budget value among the reported methods on four synthetic families and both engineering benchmarks. These results indicate that sparse records from completed related tasks can support adaptive query decisions on new tasks, making more effective use of limited, costly experimental evaluations.
Abstract:In experiment-driven black-box optimization, each evaluation can consume materials, equipment time, and other resources, limiting the number of experiments available for a new task. Records from completed related optimization campaigns could help, but learning a sequential query policy remains challenging when task-specific experience consists of only a few sparse histories. We propose CAP (Candidate Acquisition Policy), which models shared structure and task-specific variation in fixed historical records to sample interactive practice functions. On these functions, CAP learns a feedback-conditioned query policy aimed at final optimization performance under a fixed evaluation budget. At deployment, its parameters remain fixed while each new observation informs the next query. Across five synthetic task families spanning 6 and 20 dimensions and two engineering simulation benchmarks, BOPTEST and SU2, CAP attains the lowest median final-budget value among the reported methods on four synthetic families and both engineering benchmarks. These results indicate that sparse records from completed related tasks can support adaptive query decisions on new tasks, making more effective use of limited, costly experimental evaluations.
Machine Learning (cs.LG)
arXiv:2610.41173 [cs.LG]
(or arXiv:2610.41173v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.41173 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 263 words
Related coverage
Can you see how I learn? Human observers' inferences about Reinforcement Learning agents' learning processes
arXiv · Aug 24, 2026
Toward Learning POMDPs Beyond Full-Rank Actions and State Observability
arXiv · Jul 23, 2026
Observation-Aligned Mask Priors for Learning Physical Fields from Authentic Occlusions
arXiv · Sep 30, 2026
Online World Modeling Enables Real-World Inverse Reinforcement Learning from Observation
arXiv · Oct 20, 2026