Skip to content

Dataset · PKU-Alignment

PKU-SafeRLHF safety and alignment data

PKU-Alignment/PKU-SafeRLHF-10K

Tabular and text corpus for training and evaluating language model safety using RLHF.

What's inside

An English-language dataset containing 10k to 100k entries in tabular and text formats. It is intended for text generation tasks and is associated with a research paper (arXiv:2307.04657) on model alignment.

What you can do with it

  • Training reward models for RLHF taking response safety into account
  • SFT fine-tuning or alignment of LLMs to reduce toxicity and harm
Tasks
Text generation
Languages
English
Size
10K–100K rows
Downloads this month
1.5K
Likes
63

How to load it

from datasets import load_dataset

ds = load_dataset("PKU-Alignment/PKU-SafeRLHF-10K")