Llm-driven AutoML with trial correction boosts reliability and performance
SAGE-Loop: Reliable Closed-Loop LLM-Driven AutoML with Trial-and-Correction and Adaptive Ensembling
Machine Learning
Summary
AutoML is a way to automatically build machine learning models, but it often struggles when things go wrong during the process, causing poor results or failures. The authors created SAGE-Loop, a new system that uses large language models to repeatedly generate and check solutions, fix errors, and smartly combine different models to improve reliability. This approach helps AutoML recover from mistakes and perform better on a variety of tasks like classification, regression, and clustering. The system was tested on many datasets and showed consistent improvements in how well and how stably it works.
What this means in practice
- •For machine learning engineers: Build more reliable and adaptive AutoML pipelines that recover from failures and improve model combinations automatically.
- •For data platform teams: Maintain stable machine learning workflows in industrial settings with automated trial correction and ensemble adaptation across diverse tasks.
Authors
Junquan Gu, Shibo Cui, Xiangfeng Luo, Hang Yu
Abstract
Automated machine learning (AutoML) is reshaping data-driven science and industrial practice, and as large language models are introduced into AutoML, pipeline reliability becomes as important as automation efficiency. However, existing AutoML still struggles to realize instant feedback and adaptive optimization during execution, so once a run drifts into a suboptimal or failed state, it lacks a process-level correction mechanism. The fundamental pathology lies in its one-way pipeline: intermediate failures are typically terminated or bypassed, while fixed paradigms often strengthen model generation but leave ensemble decisions static, weakening both execution reliability and the controlled use of structural diversity. This indicates that LLM-driven AutoML needs a closed-loop ability for trial-correction-improvement together with evidence-based use of model diversity. To this end, we propose SAGE-Loop, a reliable closed-loop, self-adaptive, LLM-driven AutoML framework that performs multi-round generation and validation for trial-and-repair, and adaptively selects ensemble strategies in both supervised and unsupervised tasks, thereby unifying how to generate with how to use models. Across 20 public datasets, SAGE-Loop consistently improves performance and stability on classification, regression, and clustering tasks. Additional results further show its ability to recover from execution failures and maintain robust pipeline behavior.