AutoLLMResearch: Training Research Agents for Automating LLM Experiment Configuration -- Learning from Cheap, Optimizing Expensive

2026-05-12

AI Summary

This work addresses the challenge of automating experimental configuration for high-cost large language models (LLMs), a process traditionally reliant on expert intuition and lacking efficient methods for navigating expensive configuration spaces. The authors propose an agent-based framework that formulates configuration optimization as a long-horizon Markov decision process, leveraging multi-fidelity environments to learn generalizable patterns from low-cost experiments and extrapolate them to high-cost LLM settings. To enable this approach, they introduce LLMConfig-Gym, a benchmark environment encompassing millions of GPU hours of training data, along with a structured training protocol that facilitates cross-fidelity optimization. Empirical results demonstrate that the proposed method significantly outperforms strong baselines across multiple unseen tasks, exhibiting exceptional generalization, effectiveness, and interpretability—marking the first successful realization of automated configuration for high-cost LLM experimentation.

Abstract

Effectively configuring scalable large language model (LLM) experiments, spanning architecture design, hyperparameter tuning, and beyond, is crucial for advancing LLM research, as poor configuration choices can waste substantial computational resources and prevent models from realizing their full potential. Prior automated methods are designed for low-cost settings where repeated trial and error is feasible, but scalable LLM experiments are too expensive for such extensive iteration. To our knowledge, no work has addressed the automation of high-cost LLM experiment configurations, leaving this problem labor-intensive and dependent on expert intuition. Motivated by this gap, we propose AutoLLMResearch, an agentic framework that mimics how human researchers learn generalizable principles from low-fidelity experiments and extrapolate to efficiently identify promising configurations in expensive LLM settings. The core challenge is how to enable an agent to learn, through interaction with a multi-fidelity experimental environment that captures the structure of the LLM configuration landscape. To achieve this, we propose a systematic framework with two key components: 1) LLMConfig-Gym, a multi-fidelity environment encompassing four critical LLM experiment tasks, supported by over one million GPU hours of verifiable experiment outcomes; 2) A structured training pipeline that formulates configuration research as a long-horizon Markov Decision Process and accordingly incentivizes cross-fidelity extrapolation reasoning. Extensive evaluation against diverse strong baselines on held-out experiments demonstrates the effectiveness, generalization, and interpretability of our framework, supporting its potential as a practical and general solution for scalable real-world LLM experiment automation.

Problem

• Large Language Models
• Experiment Configuration
• Automation
• Multi-fidelity Optimization
• Hyperparameter Tuning

Innovation

• AutoLLMResearch
• multi-fidelity optimization
• LLM experiment automation
• cross-fidelity extrapolation
• Markov Decision Process