SMCEvolve: Principled Scientific Discovery via Sequential Monte Carlo Evolution
The paper introduces SMCEvolve, a new framework for automated scientific discovery using Sequential Monte Carlo methods. It emphasizes principled design components that enhance program evolution efficiency. The authors claim that SMCEvolve outperforms existing systems while minimizing the number of required LLM calls.
- ▪SMCEvolve recasts program search as sampling from a reward-tilted target distribution.
- ▪The framework includes adaptive parent resampling, mixture of mutation with acceptance, and automatic convergence control as core mechanisms.
- ▪SMCEvolve demonstrates superior performance across various benchmarks, including math and machine learning tasks.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.15308 (cs) [Submitted on 14 May 2026] Title:SMCEvolve: Principled Scientific Discovery via Sequential Monte Carlo Evolution Authors:Jiachen Jiang, Huminhao Zhu, Zhihui Zhu View a PDF of the paper titled SMCEvolve: Principled Scientific Discovery via Sequential Monte Carlo Evolution, by Jiachen Jiang and Huminhao Zhu and Zhihui Zhu View PDF HTML (experimental) Abstract:LLM-driven program evolution has emerged as a powerful tool for automated scientific discovery, yet existing frameworks offer no principled guide for designing their individual components and provide no guarantee that the search converges.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
