Search-Based Software Engineering and AI Foundation Models: Current Landscape and Future Roadmap

This research roadmap examines the evolving relationship between search-based software engineering (SBSE) and AI foundation models like large language models, after 25 years of SBSE development. The paper identifies three core integration pathways: using FMs to enhance SBSE techniques, applying SBSE methods to improve FM development, and exploring synergies between both approaches for future software engineering challenges.

Search-based software engineering represents a mature research discipline that applies metaheuristic optimization techniques to solve complex software engineering problems across the entire development lifecycle. The emergence of foundation models, particularly large language models, creates a critical inflection point where established SBSE methodologies must adapt to leverage new AI capabilities while potentially offering complementary strengths to FM development.

The research landscape has evolved significantly over two and a half decades, with SBSE proving effective across diverse domains including test generation, program repair, and architectural optimization. Foundation models introduce unprecedented capabilities for code understanding, generation, and analysis, yet operate under different paradigms than traditional search-based approaches. This creates both displacement risk and integration opportunity—FMs could automate tasks previously handled by SBSE techniques, or the two approaches could combine to exceed individual capabilities.

For software engineering teams and tool developers, this convergence carries immediate implications. Organizations currently investing in SBSE-based solutions face potential disruption as FM-based alternatives emerge, while those adopting only FM approaches may miss optimization benefits from search techniques. The research roadmap addresses practical questions: Can FMs serve as fitness evaluators or solution generators within SBSE frameworks? Can SBSE techniques improve FM training efficiency or safety? Can hybrid architectures outperform either approach independently?

The emerging field will likely consolidate around hybrid methodologies that combine FM flexibility with SBSE's principled optimization guarantees. Teams should monitor this convergence closely, as architectural decisions made today regarding testing, debugging, and code generation tools may require substantial revision within 12-24 months as standards emerge.

• →Foundation models and search-based software engineering represent complementary paradigms that can integrate across three dimensions: FMs enhancing SBSE, SBSE improving FM development, and hybrid synergistic approaches.
• →The 25-year SBSE discipline faces both disruption from LLM automation and opportunity through collaborative integration with FM capabilities.
• →Practical applications span using FMs as fitness evaluators, SBSE optimization of model training, and hybrid architectures for software engineering tasks.
• →Organizations must reassess existing SBSE tool investments against emerging FM-based alternatives while exploring integration rather than replacement strategies.
• →Hybrid approaches combining FM flexibility with SBSE's principled optimization likely represent the future standard for advanced software engineering automation.