ADAPT for Science: A Multi-Agent System for Self-Evolving Scientific Optimization
Abstract
Scientific discovery often requires identifying a small number of high-performing designs in vast, constrained search spaces. This is difficult when evaluations are costly, feasible solutions are sparse, parameters are strongly coupled, and physical constraints are strict. Recent AI for AI methods can automatically improve algorithms and optimization procedures. However, general purpose approaches are not designed around scientific priors, physical constraints, and limited evaluation budgets. We introduce ADAPT for Science, a multi-agent framework that treats scientific optimization as two coupled decisions: where to search and how to search. ADAPT uses historical designs to identify promising and diverse initial conditions, progressively concentrates computation on favorable search directions, and evolves the optimization procedure from execution feedback. We evaluate ADAPT on the design of gravitational wave detectors using a differentiable physics simulator and eight unseen detector topologies. Under matched scientific evaluation budgets, ADAPT achieves a mean loss of 0.204 +/- 0.101, reducing loss by 54.9% relative to the best classical optimizer and by 62.2% relative to the best general purpose agentic evolution baseline. ADAPT achieves the lowest loss on all unseen topologies among the evaluated baselines. ADAPT provides a step toward scientific optimization systems that continuously learn from prior experience and their own search outcomes.
Related
- Self-Evolving Scientific Agent Designs Physically Reasoned White-Box Fluid Control
- Agentic programs: an emerging form of scientific software in computational materials science
- Mined from Scientific Literature: Process Schemas for Atomic Layer Deposition and Etching in Materials Science
- False-science induction in autonomous scientific discovery
- Human-in-the-Loop Meta Bayesian Optimization for Fusion Energy and Scientific Applications
BibTeX
@misc{feng2026adapt,
title = {{ADAPT for Science: A Multi-Agent System for Self-Evolving Scientific Optimization}},
author = {Zhaoran Feng and Yuchen Ni and Jianan Liu and Lemao Liu and Dejing Dou and Kairui Feng},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.30851},
eprint = {2610.30851},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.30851},
}
RIS
TY - UNPB
TI - ADAPT for Science: A Multi-Agent System for Self-Evolving Scientific Optimization
AU - Zhaoran Feng
AU - Yuchen Ni
AU - Jianan Liu
AU - Lemao Liu
AU - Dejing Dou
AU - Kairui Feng
PY - 2026
DA - 2026/10/22
PB - arXiv
DO - 10.48550/arxiv.2610.30851
UR - https://arxiv.org/abs/2610.30851
AB - Scientific discovery often requires identifying a small number of high-performing designs in vast, constrained search spaces. This is difficult when evaluations are costly, feasible solutions are sparse, parameters are strongly coupled, and physical constraints are strict. Recent AI for AI methods can automatically improve algorithms and optimization procedures. However, general purpose approaches are not designed around scientific priors, physical constraints, and limited evaluation budgets. We introduce ADAPT for Science, a multi-agent framework that treats scientific optimization as two coupled decisions: where to search and how to search. ADAPT uses historical designs to identify promising and diverse initial conditions, progressively concentrates computation on favorable search directions, and evolves the optimization procedure from execution feedback. We evaluate ADAPT on the design of gravitational wave detectors using a differentiable physics simulator and eight unseen detector topologies. Under matched scientific evaluation budgets, ADAPT achieves a mean loss of 0.204 +/- 0.101, reducing loss by 54.9% relative to the best classical optimizer and by 62.2% relative to the best general purpose agentic evolution baseline. ADAPT achieves the lowest loss on all unseen topologies among the evaluated baselines. ADAPT provides a step toward scientific optimization systems that continuously learn from prior experience and their own search outcomes.
ER -