SeaEvo Evolutionary Loop Framework Overview
Figure 2 : Overview of SeaEvo . Three modules enrich the evolutionary loop: Strategy Articulation replaces blind mutation with a diagnose-direct-implement pipeline; Stratified Experience Retrieval assembles context from semantically clustered, behaviorally diverse archive entries; Strategic Landscape Navigation periodically diagnoses which strategy families are effective, saturated, or underexplored, and steers future mutations accordingly.
논문 컨텍스트
Paper title: SeaEvo: Advancing Algorithm Discovery with Strategy Space Evolution Abstract: LLM-guided evolutionary search has emerged as a promising paradigm for automated algorithm discovery, yet most systems track search progress primarily through executable programs and scalar fitness. Even when natural-language reflection is used, it is often used locally in mutation prompts or stored without an explicit population-level organization of strategic directions. As a result, evolutionary search can struggle to distinguish syntactically different implementations of the same idea, preserve lower-fitness but strategically promising directions, or detect when an entire family of strategies has saturated. We introduce \model, a modular strategy-space layer that elevates natural-language strategy descriptions from transient prompt context to first-class population-level evolutionary state in LLM-driven program search. \model augments each candidate program with an explicit natural language strategy description and uses this representation in three ways: Strategy Articulation turns mutation into a diagnose-direct-implement process; Stratified Experience Retrieval organizes the archive into strategy clusters and selects inspirations by behavioral complementarity; and Strategic Landscape Navigation periodically summarizes effective, saturated, and underexplored strategy families to guide future mutations. Across mathematical algorithm discovery, systems optimization, and agent-scaffold benchmarks, \model improves the underlying evolutionary backbones in most settings, wit Passages referencing this figure: efine candidate programs, achieving strong results across mathematical discovery, combinatorial and geometric optimization, systems engineering, and agentic program design (Liu et al. , 2024 ; Ye et al. , 2024 ; Liu et al. , 2026a ) . This paradigm is especially attractive in domains where search spaces are combinatorially large and manually engineered heuristics are costly to design or transfer. Figure 1 : Search trajectories on Circle Packing benchmarks. Solid curves show the best-so-far sum of radii over generations, dotted curves show cumulative API cost, and the gray dashed line denotes the Human/SOTA reference. SeaEvo improves both OpenEvolve and ShinkaEvolve backbones, reaching competitive solutions earlier while incurring lower cumulative cost. Despite this progress, most LLM-drive val Δ = 10 \Delta=10 , C = 5 C=5 strategy clusters, and ε \varepsilon -greedy ratio ε = 0.2 \varepsilon=0.2 . Each experiment is repeated three times; we report mean and standard deviation. More details are provided in Appendix B . 4.2 Main Results Tables 1 and 2 report results on mathematical and systems optimization benchmarks, respectively. Figure 4 reports agent-scaffold design on XSTest, and Figure 1 shows search trajectories on the two Circle Packing benchmarks. We highlight several observations. Consistent improvement across tasks and backbones. Across most configurations in Tables 1 and 2 , SeaEvo achieves the highest average fitness, and no baseline systematically dominates on any task or backbone. The relative ranking is stable between MiMo-V2-Pro and Gemini-3-Flash, suggesting t 4.3 % +4.3\% on Circle Packing Square Avg in Table 1 ), and diminish on narrower-solution-space tasks such as LLM-SQL. We attribute this gradient to SLN and SER providing stronger signal when qualit
