AutoAdapt: Microsoft's New Framework Automates LLM Domain Adaptation

Microsoft Research has introduced AutoAdapt, a framework designed to automate domain adaptation for large language models (LLMs). Instead of relying on manual trial and error, AutoAdapt uses an Adaptation Configuration Graph (ACG) to map optimal adaptation paths and a Planning Agent to select between strategies like retrieval-augmented generation (RAG) and fine-tuning. The system also includes AutoRefine, which optimizes hyperparameters within specified budget and hardware constraints, reducing weeks of manual iteration into a repeatable pipeline. This approach aims to make domain-specific LLM deployment more efficient and reproducible across fields such as law, medicine, and tech support.