Pile of Law: Learning Responsible Data Filtering from the Law and a 256GB Open-Source Legal Dataset

TL;DR

This paper introduces the Pile of Law, a large open-source legal dataset, and proposes a law-based filtering approach to improve responsible data curation for large language models, aiming to reduce harmful content.

Contribution

It presents a new legal dataset and a law-grounded filtering methodology, enabling models to learn content restrictions directly from data, addressing ethical concerns in pretraining.

Findings

The Pile of Law dataset contains 256GB of legal and administrative data.

Legal norms can be distilled into actionable filtering lessons.

Researchers can learn filtering rules directly from the dataset.

Abstract

One concern with the rise of large language models lies with their potential for significant harm, particularly from pretraining on biased, obscene, copyrighted, and private information. Emerging ethical approaches have attempted to filter pretraining material, but such approaches have been ad hoc and failed to take context into account. We offer an approach to filtering grounded in law, which has directly addressed the tradeoffs in filtering material. First, we gather and make available the Pile of Law, a 256GB (and growing) dataset of open-source English-language legal and administrative data, covering court opinions, contracts, administrative rules, and legislative records. Pretraining on the Pile of Law may help with legal tasks that have the promise to improve access to justice. Second, we distill the legal norms that governments have developed to constrain the inclusion of toxic or private content into actionable lessons for researchers and discuss how our dataset reflects these norms. Third, we show how the Pile of Law offers researchers the opportunity to learn such filtering rules directly from the data, providing an exciting new research direction in model-based processing.