論文 arXiv 発表: 2026-06-04

MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery

MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery

著者: Shangheng Du, Xiangchao Yan, Jinxin Shi, Zongsheng Cao, Shiyang Feng ほか9名

要約

Large language model (LLM) agents are increasingly applied to long-horizon tasks such as scientific discovery and machine learning engineering (MLE), where sustained self-evolution becomes a key capability. However, existing MLE agents suffer from inter-branch information isolation, memoryless searc…

#agent#llm#coding#benchmark

同じカテゴリの記事