arXiv 2506.15692
MLE-STAR: Machine Learning Engineering Agent via Search and Targeted Refinement
By Jaehyun Nam, Jinsung Yoon, et al.
Published 2025-05-27
Citation lineage
Review the prior work and downstream research connected to this paper.
Agents based on large language models (LLMs) for machine learning engineering (MLE) can automatically implement ML models via code generation. However, existing approaches to build such agents often rely heavily on inherent LLM knowledge and employ coarse exploration strategies that modify the entire code structure at once. This limits their ability to select effective task-specific models and perform deep explorati…