extract 0 0 0

lvogel123/jailbreak-gpt-5-high

published by @lvogel123

Alignment dataset tracked in the /datasets sub-catalog.

Lifetime downloads
181
Last 30 days
39
Likes
-
Size
64 KB
Created on HF
2025-10-25
Age
12 months ago

Tags

size_categories:n<1Kformat:parquetformat:optimized-parquetmodality:tabularmodality:textlibrary:datasetslibrary:pandaslibrary:polarslibrary:mlcroissantregion:us

README current version from Hugging Face


dataset_info:

  • config_name: mhj_m2s_gpt_5_20251024-235542_results
    features:
    • name: log_path
      dtype: string
    • name: eval_id
      dtype: string
    • name: run_id
      dtype: string
    • name: created
      dtype: string
    • name: task
      dtype: string
    • name: task_id
      dtype: string
    • name: model
      dtype: string
    • name: total_samples
      dtype: int64
    • name: completed_samples
      dtype: int64
    • name: mean
      dtype: float64
    • name: score01_mean
      dtype: float64
    • name: [email protected]
      dtype: float64
    • name: [email protected]
      dtype: float64
    • name: refusal_rate
      dtype: float64
    • name: convincing_raw_mean
      dtype: float64
    • name: specificity_raw_mean
      dtype: float64
    • name: convincing01_mean
      dtype: float64
    • name: specificity01_mean
      dtype: float64
    • name: raw_mean
      dtype: float64
    • name: Physical Harm & Weapons score01_mean
      dtype: float64
    • name: Physical Harm & Weapons [email protected]
      dtype: float64
    • name: Physical Harm & Weapons [email protected]
      dtype: float64
    • name: Physical Harm & Weapons refusal_rate
      dtype: float64
    • name: Physical Harm & Weapons convincing_raw_mean
      dtype: float64
    • name: Physical Harm & Weapons specificity_raw_mean
      dtype: float64
    • name: Financial Crime & Fraud score01_mean
      dtype: float64
    • name: Financial Crime & Fraud [email protected]
      dtype: float64
    • name: Financial Crime & Fraud [email protected]
      dtype: float64
    • name: Financial Crime & Fraud refusal_rate
      dtype: float64
    • name: Financial Crime & Fraud convincing_raw_mean
      dtype: float64
    • name: Financial Crime & Fraud specificity_raw_mean
      dtype: float64
    • name: IP/Copyright & Non-Violent Unethical Behavior score01_mean
      dtype: float64
    • name: IP/Copyright & Non-Violent Unethical Behavior [email protected]
      dtype: float64
    • name: IP/Copyright & Non-Violent Unethical Behavior [email protected]
      dtype: float64
    • name: IP/Copyright & Non-Violent Unethical Behavior refusal_rate
      dtype: float64
    • name: IP/Copyright & Non-Violent Unethical Behavior convincing_raw_mean
      dtype: float64
    • name: IP/Copyright & Non-Violent Unethical Behavior specificity_raw_mean
      dtype: float64
    • name: Sexual Content & Exploitation score01_mean
      dtype: float64
    • name: Sexual Content & Exploitation [email protected]
      dtype: float64
    • name: Sexual Content & Exploitation [email protected]
      dtype: float64
    • name: Sexual Content & Exploitation refusal_rate
      dtype: float64
    • name: Sexual Content & Exploitation convincing_raw_mean
      dtype: float64
    • name: Sexual Content & Exploitation specificity_raw_mean
      dtype: float64
    • name: Drugs & Controlled Substances score01_mean
      dtype: float64
    • name: Drugs & Controlled Substances [email protected]
      dtype: float64
    • name: Drugs & Controlled Substances [email protected]
      dtype: float64
    • name: Drugs & Controlled Substances refusal_rate
      dtype: float64
    • name: Drugs & Controlled Substances convincing_raw_mean
      dtype: float64
    • name: Drugs & Controlled Substances specificity_raw_mean
      dtype: float64
    • name: Hate/Harassment/Abuse score01_mean
      dtype: float64
    • name: Hate/Harassment/Abuse [email protected]
      dtype: float64
    • name: Hate/Harassment/Abuse [email protected]
      dtype: float64
    • name: Hate/Harassment/Abuse refusal_rate
      dtype: float64
    • name: Hate/Harassment/Abuse convincing_raw_mean
      dtype: float64
    • name: Hate/Harassment/Abuse specificity_raw_mean
      dtype: float64
    • name: Misinformation score01_mean
      dtype: float64
    • name: Misinformation [email protected]
      dtype: float64
    • name: Misinformation [email protected]
      dtype: float64
    • name: Misinformation refusal_rate
      dtype: float64
    • name: Misinformation convincing_raw_mean
      dtype: float64
    • name: Misinformation specificity_raw_mean
      dtype: float64
    • name: Chemical/Biological Threats score01_mean
      dtype: float64
    • name: Chemical/Biological Threats [email protected]
      dtype: float64
    • name: Chemical/Biological Threats [email protected]
      dtype: float64
    • name: Chemical/Biological Threats refusal_rate
      dtype: float64
    • name: Chemical/Biological Threats convincing_raw_mean
      dtype: float64
    • name: Chemical/Biological Threats specificity_raw_mean
      dtype: float64
    • name: Self-Harm & Suicide score01_mean
      dtype: float64
    • name: Self-Harm & Suicide [email protected]
      dtype: float64
    • name: Self-Harm & Suicide [email protected]
      dtype: float64
    • name: Self-Harm & Suicide refusal_rate
      dtype: float64
    • name: Self-Harm & Suicide convincing_raw_mean
      dtype: float64
    • name: Self-Harm & Suicide specificity_raw_mean
      dtype: float64
    • name: Cybercrime & Intrusion score01_mean
      dtype: float64
    • name: Cybercrime & Intrusion [email protected]
      dtype: float64
    • name: Cybercrime & Intrusion [email protected]
      dtype: float64
    • name: Cybercrime & Intrusion refusal_rate
      dtype: float64
    • name: Cybercrime & Intrusion convincing_raw_mean
      dtype: float64
    • name: Cybercrime & Intrusion specificity_raw_mean
      dtype: float64
      splits:
    • name: train
      num_bytes: 12689
      num_examples: 15
      download_size: 48688
      dataset_size: 12689
  • config_name: mhj_m2s_gpt_5_20251024-235542_stats
    features:
    • name: log_path
      dtype: string
    • name: eval_id
      dtype: string
    • name: run_id
      dtype: string
    • name: created
      dtype: string
    • name: task
      dtype: string
    • name: task_id
      dtype: string
    • name: model
      dtype: string
    • name: started_at
      dtype: string
    • name: completed_at
      dtype: string
    • name: usage_model
      dtype: string
    • name: input_tokens
      dtype: int64
    • name: output_tokens
      dtype: int64
    • name: total_tokens
      dtype: int64
      splits:
    • name: train
      num_bytes: 4008
      num_examples: 11
      download_size: 8841
      dataset_size: 4008
      configs:
  • config_name: mhj_m2s_gpt_5_20251024-235542_results
    data_files:
    • split: train
      path: mhj_m2s_gpt_5_20251024-235542_results/train-*
  • config_name: mhj_m2s_gpt_5_20251024-235542_stats
    data_files:
    • split: train
      path: mhj_m2s_gpt_5_20251024-235542_stats/train-*

README history 3 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. 2025-10-25Upload datasetebf0f795.9 KB
    Loading...
  2. 2025-10-25Upload dataset86ea6455.1 KB
    Loading...
  3. 2025-10-25initial commit6560aa40 B
    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.

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("lvogel123/jailbreak-gpt-5-high")
Raw snapshot (Python)
from huggingface_hub import snapshot_download
snapshot_download(repo_id="lvogel123/jailbreak-gpt-5-high", repo_type="dataset")
git clone (requires git-lfs)
git clone https://huggingface.co/datasets/lvogel123/jailbreak-gpt-5-high
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