{
 "schema_version": 1,
 "generated": "2026-10-20",
 "repo": "p-e-w/heretic",
 "name": "heretic",
 "owner": "p-e-w",
 "slug": "p-e-w-heretic",
 "github": "https://github.com/p-e-w/heretic",
 "what": "Heretic is a command-line tool that automatically removes safety alignment (censorship) from transformer language models using directional ablation tuned by Optuna.",
 "why": "",
 "language": "Python",
 "requirements": [
  "Python 3.10 or newer",
  "PyTorch 2.2 or newer, installed for your hardware",
  "A GPU with enough VRAM for the model you pick (the README cites an RTX 3090 and an RTX 5090)",
  "Access to the target model's files, such as a Hugging Face model ID"
 ],
 "audience_note": "Written for ML practitioners and local-model users who can run a Python command-line tool on a GPU machine.",
 "risk": "dual-use",
 "section": "earlier",
 "weekly_rank": null,
 "category": "Research",
 "url": "https://dedupedai.com/use-cases/p-e-w-heretic/",
 "use_case_count": 4,
 "use_cases": [
  {
   "id": "p-e-w-heretic--decensor-local-open-model",
   "title": "Remove refusals from an open model in one run",
   "best": true,
   "url": "https://dedupedai.com/use-cases/p-e-w-heretic/decensor-local-open-model/",
   "json": "https://dedupedai.com/use-cases/u/p-e-w-heretic/decensor-local-open-model.json"
  },
  {
   "id": "p-e-w-heretic--verify-decensored-model-against-original",
   "title": "Measure refusals and KL divergence against the original model",
   "best": false,
   "url": "https://dedupedai.com/use-cases/p-e-w-heretic/verify-decensored-model-against-original/",
   "json": "https://dedupedai.com/use-cases/u/p-e-w-heretic/verify-decensored-model-against-original.json"
  },
  {
   "id": "p-e-w-heretic--fit-decensoring-into-limited-vram",
   "title": "Decensor a larger model on a small GPU with 4-bit quantization",
   "best": false,
   "url": "https://dedupedai.com/use-cases/p-e-w-heretic/fit-decensoring-into-limited-vram/",
   "json": "https://dedupedai.com/use-cases/u/p-e-w-heretic/fit-decensoring-into-limited-vram.json"
  },
  {
   "id": "p-e-w-heretic--study-directional-ablation-internals",
   "title": "Trace how refusal directions are ablated across layers",
   "best": false,
   "url": "https://dedupedai.com/use-cases/p-e-w-heretic/study-directional-ablation-internals/",
   "json": "https://dedupedai.com/use-cases/u/p-e-w-heretic/study-directional-ablation-internals.json"
  }
 ],
 "disclaimer": "Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk."
}