AutoAdapt: An Automated Domain Adaptation Framework for LLMs

TL;DR

AutoAdapt is an automated framework that enhances large language model domain adaptation by reducing manual effort, optimizing hyperparameters efficiently, and leveraging knowledge bases, leading to significant accuracy improvements across multiple tasks.

Contribution

The paper introduces AutoAdapt, a novel end-to-end automated framework utilizing a multi-agent debating system and AutoRefine for hyperparameter optimization in LLM domain adaptation.

Findings

AutoAdapt achieves 25% average accuracy improvement over SOTA baselines.

The multi-agent debating system effectively aligns user intent and data signals.

AutoRefine reduces hyperparameter tuning costs significantly.

Abstract

Large language models (LLMs) excel in open domains but struggle in specialized settings with limited data and evolving knowledge. Existing domain adaptation practices rely heavily on manual trial-and-error processes, incur significant hyperparameter complexity, and are highly sensitive to data and user preferences, all under the high cost of LLM training. Moreover, the interactions and transferability of hyperparameter choices across models/domains remain poorly understood, making adaptation gains uncertain even with substantial effort. To solve these challenges, we present AutoAdapt, a novel end-to-end automated framework for efficient and reliable LLM domain adaptation. AutoAdapt leverages curated knowledge bases from literature and open-source resources to reduce expert intervention. To narrow the search space, we design a novel multi-agent debating system in which proposal and critic agents iteratively interact to align user intent and incorporate data signals and best practices into the planning process. To optimize hyperparameters under tight budgets, we propose AutoRefine, a novel LLM-based surrogate that replaces costly black-box search. Across 10 tasks, AutoAdapt achieves a 25% average relative accuracy improvement over state-of-the-art Automated Machine Learning baselines with minimal overhead.