← 返回论文库

AlphaEvolve: A Gemini-Powered Coding Agent for Designing Advanced Algorithms

DeepMind · 2025
L1 · Domain Research AgentsarXiv:2506.13131#program-search#evolutionary
CORE IDEA
FunSearch 升级版：演化长程序，Gemini Flash+Pro 双模型分工，多 evaluator Pareto 优化。
CONCRETE EXAMPLE
4×4 复数矩阵乘法从 49 次降到 48 次（57 年来首次改进）。
L-ANCHOR · 为什么在这一层重要
FunSearch 工程化，最接近 production 的 AI for algorithm discovery
arXiv:2506.13131 ↗源码 ↗

相关论文

QuantaAlpha: An Evolutionary Framework for LLM-Driven Alpha Mining
L0.32026
FunSearch: Mathematical Discoveries from Program Search with LLMs
L12023
AI Scientist v1: Towards Fully Automated Open-Ended Scientific Discovery
L12024
AI Scientist v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search
L12025