extract 0 0 0

iagoalves/jailbreaking_toxsyn_gemma_llama_qwen_class

published by @iagoalves

Alignment dataset tracked in the /datasets sub-catalog.

Lifetime downloads
300
Last 30 days
13
Likes
-
Size
216.5 MB
Created on HF
2025-11-17
Age
11 months ago

Tags

size_categories:1K<n<10Kformat:parquetmodality:tabularmodality:textlibrary:datasetslibrary:pandaslibrary:mlcroissantlibrary:polarsregion:us

README current version from Hugging Face


dataset_info:
features:

  • name: id
    dtype: int64
  • name: text
    dtype: string
  • name: prompt_strategy
    dtype: string
  • name: is_toxic
    dtype: int64
  • name: group
    dtype: string
  • name: scenario_theme
    dtype: string
  • name: scenario
    dtype: string
  • name: jb_strategy
    dtype: string
  • name: jb_prompt
    dtype: string
  • name: jb_response
    dtype: string
  • name: jb_class
    dtype: string
  • name: jb_response_gemma-3-1b-it
    dtype: string
  • name: jb_response_gemma-3-4b-it
    dtype: string
  • name: jb_response_gemma-3-12b-it
    dtype: string
  • name: jb_response_gemma-3-27b-it
    dtype: string
  • name: jb_response_Llama-3.2-1B-Instruct
    dtype: string
  • name: jb_response_Llama-3.2-3B-Instruct
    dtype: string
  • name: jb_response_Llama-3.1-8B-Instruct
    dtype: string
  • name: jb_response_Qwen3-1.7B
    dtype: string
  • name: jb_response_Qwen3-4B
    dtype: string
  • name: jb_response_Qwen3-8B
    dtype: string
  • name: jb_response_Qwen3-32B
    dtype: string
  • name: jb_response_Llama-3.3-70B-Instruct
    dtype: string
  • name: jb_class_Llama-3.1-8B-Instruct
    dtype: string
  • name: jb_class_Llama-3.3-70B-Instruct
    dtype: string
  • name: jb_class_Qwen3-1.7B
    dtype: string
  • name: jb_class_Qwen3-4B
    dtype: string
  • name: jb_class_Qwen3-8B
    dtype: string
  • name: jb_class_Qwen3-32B
    dtype: string
  • name: jb_class_Llama-3.2-1B-Instruct
    dtype: string
  • name: jb_class_Llama-3.2-3B-Instruct
    dtype: string
  • name: jb_class_gemma-3-1b-it
    dtype: string
  • name: jb_class_gemma-3-4b-it
    dtype: string
  • name: jb_class_gemma-3-12b-it
    dtype: string
  • name: jb_class_gemma-3-27b-it
    dtype: string
  • name: class_parsed_gemma-3-27b-it
    dtype: int64
  • name: class_parsed_gemma-3-12b-it
    dtype: int64
  • name: class_parsed_gemma-3-4b-it
    dtype: int64
  • name: class_parsed_gemma-3-1b-it
    dtype: int64
  • name: class_parsed_Qwen3-32B
    dtype: int64
  • name: class_parsed_Qwen3-8B
    dtype: int64
  • name: class_parsed_Qwen3-4B
    dtype: int64
  • name: class_parsed_Qwen3-1.7B
    dtype: int64
  • name: class_parsed_Llama-3.3-70B-Instruct
    dtype: int64
  • name: class_parsed_Llama-3.1-8B-Instruct
    dtype: int64
  • name: class_parsed_Llama-3.2-3B-Instruct
    dtype: int64
  • name: class_parsed_Llama-3.2-1B-Instruct
    dtype: int64
    splits:
  • name: train
    num_bytes: 444387164
    num_examples: 9997
    download_size: 226986622
    dataset_size: 444387164
    configs:
  • config_name: default
    data_files:
    • split: train
      path: data/train-*

README history 20 revisions

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.

  1. 2025-11-27Upload dataset2db6b622.6 KB
    Loading...
  2. 2025-11-27Upload dataset30555941.5 KB
    Loading...
  3. 2025-11-18Upload datasete9c193e2.6 KB
    Loading...
  4. 2025-11-18Upload dataseta6ecc972.3 KB
    Loading...
  5. 2025-11-17Upload datasetbb2e1ba2.1 KB
    Loading...
  6. 2025-11-17Upload datasetcdc82711.9 KB
    Loading...
  7. 2025-11-17Upload dataset3c851101.6 KB
    Loading...
  8. 2025-11-17Upload dataset58ea70a1.5 KB
    Loading...
  9. 2025-11-17Upload datasetb21399e1.5 KB
    Loading...
  10. 2025-11-17Upload dataset81b1f421.5 KB
    Loading...
  11. 2025-11-17Upload dataset74862ef1.4 KB
    Loading...
  12. 2025-11-17Upload dataset65b2f911.4 KB
    Loading...
  13. 2025-11-17Upload dataset56aee3d1.5 KB
    Loading...
  14. 2025-11-17Upload datasete3851431.5 KB
    Loading...
  15. 2025-11-17Upload datasetacfc6e81.4 KB
    Loading...
  16. 2025-11-17Upload datasetbe711e51.3 KB
    Loading...
  17. 2025-11-17Upload datasetcbc86061.6 KB
    Loading...
  18. 2025-11-17Upload datasetc9500421.5 KB
    Loading...
  19. 2025-11-17Upload dataseta2b992b1.4 KB
    Loading...
  20. 2025-11-17Upload dataset4023f241.4 KB
    Loading...

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.

Community discussions 1 thread · 1 open · 1 comments

Threads posted by the community on Hugging Face: bug reports, questions about usage, pull requests on the README, license clarifications. Click any thread to expand and read the full conversation inline. Newest first.

  1. 2025-11-19[bot] Conversion to Parquetopen1 💬#1
    Loading...

Related datasets same stage · extract

Read further

Repository

Download the dataset

0

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("iagoalves/jailbreaking_toxsyn_gemma_llama_qwen_class")
Raw snapshot (Python)
from huggingface_hub import snapshot_download
snapshot_download(repo_id="iagoalves/jailbreaking_toxsyn_gemma_llama_qwen_class", repo_type="dataset")
git clone (requires git-lfs)
git clone https://huggingface.co/datasets/iagoalves/jailbreaking_toxsyn_gemma_llama_qwen_class
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