minnesotanlp/lawflow-reasoning-simulation
收藏Hugging Face2025-09-05 更新2025-07-05 收录
下载链接：
https://hf-mirror.com/datasets/minnesotanlp/lawflow-reasoning-simulation
下载链接
链接失效反馈官方服务：
资源简介：
LawFlow数据集旨在通过模拟律师的思维过程来提升大型语言模型在法律推理方面的能力，特别是实体形成任务。该数据集收集了由法律专家注释并由LLM模拟的真实场景中的推理行为。数据包括高级计划、执行细节和推理痕迹，提供了对人类和LLM推理过程的深入了解。数据集目前仅限于英语，并专注于特定的法律任务，这可能会限制其普适性。
The LawFlow dataset is designed to enhance the legal reasoning capabilities of Large Language Models, particularly focusing on entity formation tasks, by simulating lawyers thought processes. It consists of reasoning actions annotated by legal experts and simulated using LLMs on realistic scenarios. The dataset includes high-level plans, execution details, and reasoning traces, offering insights into both human and LLM reasoning processes. Currently, the dataset is limited to English and focuses on a specific legal task, which may restrict its generalizability.
提供机构：
minnesotanlp
