Alignment dataset tracked in the /datasets sub-catalog.
Original dataset size: 157093
Total unique prompts: 39948
Harmful prompts: 17845
Benign prompts: 22103
Filtered dataset size: 39925
Harmful examples: 17824
Benign examples: 22101
Prompts discarded: 23
Harmful (no refusals): 21
Benign (no complete responses): 2
Original dataset size: 157093
Filtered dataset size: 39925
Prompts discarded: 23
Every night we snapshot the README of every dataset in the catalog. When the SHA changes we archive the new version and diff it against the last. This is the evolving thought record of the alignment-data field: what the author decided to say about the corpus, and how that framing shifted over time.
No abliterated models in our catalog have declared this dataset in their YAML metadata yet. This can mean the dataset is used but not declared, is used outside abliteration workflows, or is new. The link will populate automatically as models are indexed.
Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.
from datasets import load_dataset
ds = load_dataset("Seanie-lee/ThinkSafe-R1-Distill-7B-n4-filtered-qwen")from huggingface_hub import snapshot_download
snapshot_download(repo_id="Seanie-lee/ThinkSafe-R1-Distill-7B-n4-filtered-qwen", repo_type="dataset")git clone https://huggingface.co/datasets/Seanie-lee/ThinkSafe-R1-Distill-7B-n4-filtered-qwenEvery model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.