datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
fine_grained_unlearning
Fine-Grained Knowledge Unlearning — Namesake Benchmark
A benchmark for fine-grained knowledge unlearning: can a method remove a fact
about entity X without damaging the same fact on entity Y, when X and Y
have (near-)identical names and share exactly that one attribute?
Each sample is a pair of real people who
share an identical or near-identical name,
share one career element (e.g. both are basketball players) — the fact to
unlearn on X and retain on Y,
differ on everything… See the full description on the dataset page: https://huggingface.co/datasets/ernlavr/fine_grained_unlearning.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.machine-unlearning-holdout-evals
rubenbalbastre/machine-unlearning-holdout-evals
Hold-out completions and LLM-judge rubric labels for targeted machine-unlearning
experiments.
The dataset contains 17,280 prompt/completion evaluations from
288 model runs across 10 target
entities. Each row includes the model size, reward function, training variant,
and six independent boolean rubric labels.
Rubrics
lexical_leakage: the completion mentions the target or a surface variant.
semantic_leakage: the… See the full description on the dataset page: https://huggingface.co/datasets/rubenbalbastre/machine-unlearning-holdout-evals.