TurboEvolve: Towards Fast and Robust LLM-Driven Program Evolution
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914492916432896 |
|---|---|
| author | Yang, Yang Zhong, Zining Li, Jindong Wu, Jiemin Yuan, Kaishen Chen, Wenshuo Yang, Menglin Yue, Yutao |
| author_facet | Yang, Yang Zhong, Zining Li, Jindong Wu, Jiemin Yuan, Kaishen Chen, Wenshuo Yang, Menglin Yue, Yutao |
| contents | LLM-driven program evolution can discover high-quality programs, but its cost and run-to-run variance hinder reliable progress. We propose TurboEvolve, a multi-island evolutionary framework that improves sample efficiency and robustness under fixed evaluation budgets. Inspired by the multiple-offspring strategy in evolutionary algorithms, TurboEvolve introduces verbalized Sampling, prompting the LLM to emit K diverse candidates with explicit self-assigned sampling weights, and an online scheduler that adapts K to expand exploration under stagnation and reduce overhead during steady progress. To exploit existing solution pools, we further propose "seed-pool injection," which clusters seeds and assigns them across islands with controlled perturbations and elitist preservation to balance diversity and refinement. Across multiple program-optimization benchmarks, TurboEvolve consistently achieves stronger performance at lower budgets and improves best-known solutions on several tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_18607 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | TurboEvolve: Towards Fast and Robust LLM-Driven Program Evolution Yang, Yang Zhong, Zining Li, Jindong Wu, Jiemin Yuan, Kaishen Chen, Wenshuo Yang, Menglin Yue, Yutao Neural and Evolutionary Computing Artificial Intelligence LLM-driven program evolution can discover high-quality programs, but its cost and run-to-run variance hinder reliable progress. We propose TurboEvolve, a multi-island evolutionary framework that improves sample efficiency and robustness under fixed evaluation budgets. Inspired by the multiple-offspring strategy in evolutionary algorithms, TurboEvolve introduces verbalized Sampling, prompting the LLM to emit K diverse candidates with explicit self-assigned sampling weights, and an online scheduler that adapts K to expand exploration under stagnation and reduce overhead during steady progress. To exploit existing solution pools, we further propose "seed-pool injection," which clusters seeds and assigns them across islands with controlled perturbations and elitist preservation to balance diversity and refinement. Across multiple program-optimization benchmarks, TurboEvolve consistently achieves stronger performance at lower budgets and improves best-known solutions on several tasks. |
| title | TurboEvolve: Towards Fast and Robust LLM-Driven Program Evolution |
| topic | Neural and Evolutionary Computing Artificial Intelligence |
| url | https://arxiv.org/abs/2604.18607 |