obliteratus

Remove refusal behaviors from LLMs using mechanistic interpretability techniques.

1|Updated Apr 29, 2026

Search Tags:#LLM #mechanistic interpretability #refusal removal #model surgery #AI uncensoring

Ready to run

npx skills add https://github.com/bailynlove/STARK-TOWER --skill obliteratus-bailynlove

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: obliteratus
Source: https://github.com/bailynlove/STARK-TOWER/tree/main/opencrew/skills/mlops/inference/obliteratus
Command: npx skills add https://github.com/bailynlove/STARK-TOWER --skill obliteratus-bailynlove

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires obliteratus, torch, transformers, bitsandbytes, accelerate, safetensors, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill removes refusal behaviors from large language models (LLMs), enabling uncensoring and removal of guardrails while preserving reasoning capabilities.

Core Features & Use Cases

Refusal Removal: Excise guardrails from LLMs using mechanistic interpretability techniques.
Model Ablation: Customize the removal process with various methods and parameters.
Analysis & Evaluation: Includes 28 analysis modules and built-in evaluation tools to assess model performance before and after abliteration.
Use Case: When you need to uncensor an LLM or remove refusal behaviors without retraining.

Quick Start

Run the obliteratus skill to remove refusal behaviors from a specified LLM model.

Frequently Asked Questions about obliteratus

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I remove refusal behaviors from an LLM without retraining?▼

You can remove refusal behaviors from an LLM without retraining by applying mechanistic interpretability techniques for model surgery. This process excises guardrails while preserving underlying reasoning capabilities.

What is mechanistic interpretability for AI uncensoring?▼

Mechanistic interpretability for AI uncensoring analyzes internal model activations to identify and ablate refusal directions. This enables precise model surgery to remove guardrails rather than relying on full retraining.

Do I need PyTorch and Transformers to perform model ablation?▼

Yes, you need PyTorch, Transformers, BitsAndBytes, and Accelerate installed to perform model ablation. These dependencies are required to load, analyze, and modify the LLM during the uncensoring process.

Can I evaluate LLM performance before and after abliteration?▼

Yes, you can evaluate LLM performance before and after abliteration using built-in evaluation tools. The skill includes 28 analysis modules to assess model behavior and ensure reasoning capabilities remain intact.

What is the best way to uncensor an LLM while preserving reasoning?▼

The best way to uncensor an LLM while preserving reasoning is using targeted model ablation based on mechanistic interpretability. This surgical approach removes specific refusal mechanisms without degrading overall model performance.