CoolFace
Datasetpublic

rubenbalbastre/grpo-unlearning-data

rubenbalbastre/grpo-unlearning-data Dataset splits for targeted machine unlearning experiments. License and Attribution This dataset is released as CC BY 4.0. The data is derived from RWKU (jinzhuoran/RWKU) and should be attributed to the RWKU authors. These files are a processed/modified version of RWKU: rows were filtered by forget concept, prompts were normalized, columns were renamed, splits were reorganized for this project, the reference completion column… See the full description on the dataset page: https://huggingface.co/datasets/rubenbalbastre/grpo-unlearning-data.

sourceHugging Facecc-by-4.0updated 1mo agoView on Hugging Face
0likes21downloads
Dataset Card

rubenbalbastre/grpo-unlearning-data

Dataset splits for targeted machine unlearning experiments.

License and Attribution

This dataset is released as CC BY 4.0.

The data is derived from RWKU (jinzhuoran/RWKU) and should be attributed to the RWKU authors. These files are a processed/modified version of RWKU: rows were filtered by forget concept, prompts were normalized, columns were renamed, splits were reorganized for this project, the reference completion column was removed from GRPO train and holdout splits, and a concept column was added during export.

If PURGE material is present in the source data, it is third-party material attributed to the PURGE project and is MIT-licensed by its original authors.

The source code used to create these files is licensed separately under Apache-2.0 in the machine-unlearning-llm repository.

Source dataset directory: data

Concepts:

  • Bruce Lee
  • Confucius
  • Jennifer Lopez
  • John D. Rockefeller
  • Karl Marx
  • Marlon Brando
  • Serena Williams
  • Tom Clancy
  • Tony Blair
  • Vincent van Gogh
FileSource splitRows
grpo_train.parquetgrpo/train2030
sft_train.parquetsft/train240
sft_validation.parquetsft/test130
holdout.parquetgrpo/test600