Automated Design of Graph Search Strategies for Algorithmic Optimization
Published in NeurIPS 2026 MetaAgents Workshop, 2026
We introduce the Automated Design of Graph Search Strategies (ADGSS), which replaces manual search strategy design with automated discovery in code space. Our Meta Agent Graph Search (MAGS) discovers strategies that define node selection policies, dynamic graph topologies and cross-branch information sharing, matching or outperforming hand-designed strategies on AlphaEvolve benchmark tasks.
Recommended citation: Legate, G., Deng, R., Feng, L., Qin, J., Ahmed, M. O., Tung, F. & Mori, G. (2026). "Automated Design of Graph Search Strategies for Algorithmic Optimization." NeurIPS 2026 MetaAgents Workshop.