Researchers have developed a new method for classifying environmental mitigation obligations in hydropower licensing documents. The task is typically labor-intensive and requires deep domain expertise. To address this challenge, the team introduced a Retrieval-Augmented Generation (RAG) pipeline that conditions classification on retrieved category definitions, allowing for zero-shot generalization across the full label space. This approach was evaluated on a dataset of 2017 license documents and outperformed other methods. --- Why it matters: This work matters to researchers in AI because it demonstrates a novel application of retrieval-augmented generation techniques to tackle real-world problems with limited labeled data. The proposed method has the potential to improve efficiency and accuracy in tasks that require domain expertise, such as environmental mitigation classification. Source: https://arxiv.org/abs/2608.23241