数据集 / JunkaiZ/Rubrics

JunkaiZ/Rubrics 已完整同步

Dataset Card for Dataset Name

This dataset is from the paper "Chasing the Tail: Effective Rubric-based Reward Modeling for Large Language Model Post-Training."

Dataset Details

The dataset contains 5000 training prompts and the corresponding rubrics for the health domain and the generalist domains. Each domain also includes 1000 additional prompts for the end-to-end win rate comparison. Additionally, the finance domain contains 1145 training prompts with rubrics.

Uses

The rubrics can be used for reinforcement learning from rubrics reward. For further details, please refer to the paper.

Dataset Structure

Inside the data/ directory:

  • The health/ directory contains the training set and the test set for the health domain.
  • The generalist/ directory contains the training set and the test set for the generalist domain.
  • The finance/ directory contains the training set for the finance domain.

12 个文件

浏览文件

数据集版权信息

本数据集的许可证为 MIT License。如有违反相关条款，请联系 WEHUB，我们将及时处理。 查看许可证