datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
WaterDrum-Ax
WaterDrum: Watermarking for Data-centric Unlearning Metric
WaterDrum provides an unlearning benchmark for the evaluation of the effectiveness and practicality of unlearning. The repository contains the ArXiv corpus of WaterDrum (WaterDrum-Ax), which contains both unwatermarked and watermarked ArXiv paper abstracts across
20 categories published after the release of the Llama-2 model. Each category contains 400 data samples, aggregating into 8000 samples in the full training set. The… See the full description on the dataset page: https://huggingface.co/datasets/Glow-AI/WaterDrum-Ax.WaterDrum-TOFU
WaterDrum: Watermarking for Data-centric Unlearning Metric
WaterDrum provides an unlearning benchmark for the evaluation of the effectiveness and practicality of unlearning. This repository contains the TOFU corpus of WaterDrum (WaterDrum-TOFU), which contains both unwatermarked and watermarked question-answering datasets based on the original TOFU dataset.
The data samples were watermarked with Waterfall.
Update Notice: 15/01/2026
We have updated Glow-AI/WaterDrum-TOFU to version… See the full description on the dataset page: https://huggingface.co/datasets/Glow-AI/WaterDrum-TOFU.