Bluebook任务数据集
收藏资源简介:
该数据集包含了866个Bluebook格式化任务,每个任务都有专家提供的真实答案。这些任务涵盖了从案件标题中正确缩写当事人名称,到精确使用解释性信号,再到几乎任何可能的次要来源的许多独特之处。该数据集旨在测试大型语言模型(LLMs)能否遵守美国法律程序中的一个著名复杂来源——Bluebook的规则。如果LLMs能够自动化遵守Bluebook,那么它们最终可以解放法律实践,使律师能够将资源投入到更复杂的法律推理中。
This dataset contains 866 Bluebook formatting tasks, each paired with ground-truth answers provided by subject-matter experts. These tasks span a diverse set of unique scenarios, ranging from properly abbreviating party names in case titles to accurately employing explanatory signals, and addressing nearly all potential idiosyncrasies of secondary legal sources. This dataset is designed to test whether Large Language Models (LLMs) can comply with the rules of The Bluebook, a renowned and complex authoritative source in United States legal proceedings. If LLMs can automate compliance with The Bluebook, they could ultimately transform legal practice by freeing lawyers to allocate their resources toward more complex legal reasoning.

- 1Bye-bye, Bluebook? Automating Legal Procedure with Large Language Models耶鲁大学法学院 · 2025年



