Evaluate just added

kenhktsui/test_toxic_chat

published by @kenhktsui

Safety benchmarks, over-refusal tests, and guard-model data.

size_categories:n<1Kformat:parquetmodality:tabularmodality:textlibrary:datasetslibrary:pandaslibrary:mlcroissantlibrary:polars +1 more
Lifetime downloads
585
Last 30 days
19
Likes
-
Size
95.3 KB

Description

Please note that this is subset of lmsys/toxic-chat annotated by anyclassifier.This dataset may contain racism, sexuality, or other undesired content.

Used by abliterated models none yet

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.

Metadata

Created on HF
2024-08-16 · 2 years ago
Size category
n<1K
Modalities
tabulartext
Formats
parquet
Libraries
datasetspandasmlcroissantpolars

Related datasets same stage · Evaluate

Read further

Total size
95.3 KB
Files
4
Formats
parquet
Head
213c2e0
Last updated on HF
2024-08-18

Download the dataset

Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.

Recommended: datasets library (Python)
from datasets import load_dataset
ds = load_dataset("kenhktsui/test_toxic_chat")
Raw snapshot (Python)
from huggingface_hub import snapshot_download
snapshot_download(repo_id="kenhktsui/test_toxic_chat", repo_type="dataset")
git clone (requires git-lfs)
git clone https://huggingface.co/datasets/kenhktsui/test_toxic_chat

Files by folder from our last crawl

top level 2 files 3.3 KB
.gitattributes 2.4 KB file download
README.md 863 B md download
data/ 2 files 92.0 KB
test-00000-of-00001.parquet 31.4 KB parquet download
train-00000-of-00001.parquet 60.6 KB parquet download

No README stored for this dataset yet.

Download from HuggingFace

kenhktsui/test_toxic_chat

Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every 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.

Open in Abliteration