OR-Agent: Bridging Evolutionary Search and Structured Research for Automated Algorithm Discovery
🤖 AI Summary
This work addresses the challenge that purely iterative procedural mutation struggles to effectively explore the hypothesis space in complex, experiment-driven scientific discovery. To overcome this limitation, we propose a configurable multi-agent research framework that explicitly models hypothesis branching and backtracking through a tree-structured workflow, integrating evolutionary selection with systematic exploration to enable controllable management of scientific trajectories. We introduce an innovative evolution-systematic creativity generation mechanism and design a three-tiered reflection architecture—inspired by optimization principles—that encompasses short-term experimental reflection, long-term insight accumulation, and memory compression, thereby formalizing the scientific process as a tunable dynamic system. Evaluated on combinatorial optimization and cooperative driving simulation tasks, our approach significantly outperforms strong evolutionary baselines, offering a general, scalable, and interpretable framework for AI-assisted scientific discovery.
Application Category
📝 Abstract
Automating scientific discovery in complex, experiment-driven domains requires more than iterative mutation of programs; it demands structured hypothesis management, environment interaction, and principled reflection. We present OR-Agent, a configurable multi-agent research framework designed for automated exploration in rich experimental environments. OR-Agent organizes research as a structured tree-based workflow that explicitly models branching hypothesis generation and systematic backtracking, enabling controlled management of research trajectories beyond simple mutation-crossover loops. At its core, we introduce an evolutionary-systematic ideation mechanism that unifies evolutionary selection of research starting points, comprehensive research plan generation, and coordinated exploration within a research tree. We further propose a hierarchical optimization-inspired reflection system: short-term experimental reflection operates as a form of verbal gradient providing immediate corrective signals; long-term reflection accumulates cross-experiment insights as verbal momentum; and memory compression serves as a regularization mechanism analogous to weight decay, preserving essential signals while mitigating drift. Together, these components form a principled architecture governing research dynamics. We conduct extensive experiments across classical combinatorial optimization benchmarks-including traveling salesman, capacitated vehicle routing, bin packing, orienteering, and multiple knapsack problems-as well as simulation-based cooperative driving scenarios. Results demonstrate that OR-Agent outperforms strong evolutionary baselines while providing a general, extensible, and inspectable framework for AI-assisted scientific discovery. OR-Agent source code and experiments data are publicly available at https://github.com/qiliuchn/OR-Agent.
Problem
Research questions and friction points this paper is trying to address.
automated algorithm discovery
evolutionary search
structured research
scientific discovery
hypothesis management
Innovation
Methods, ideas, or system contributions that make the work stand out.
evolutionary-systematic ideation
structured research tree
hierarchical reflection
verbal gradient
memory compression
🔎 Similar Papers
No similar papers found.
