skills.rest
NITISH-gitbit
NITISH-gitbitCOMMUNITY
Explore all skills by this author →
GitHub

obliteratus

Excise refusal directions from LLM model weights while preserving reasoning capabilities.

Updated Jun 26, 2026
Search Tags:#LLM#mechanistic interpretability#refusal removal#model surgery#ABLITERATUS
One-click install
npx skills add https://github.com/NITISH-gitbit/hermes-custom --skill obliteratus-nitish-gitbit
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: obliteratus
Source: https://github.com/NITISH-gitbit/hermes-custom/tree/main/optional-skills/mlops/obliteratus
Command: npx skills add https://github.com/NITISH-gitbit/hermes-custom --skill obliteratus-nitish-gitbit

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 allows users to uncensor and remove refusal behaviors from large language models (LLMs), such as guardrails, without retraining or fine-tuning.

Core Features & Use Cases

  • Abliterate Refusal Behaviors: Identify and surgically excise refusal directions from model weights.
  • Preserve Reasoning Capabilities: Retain the model's reasoning capabilities after removal.
  • Use Case: Use OBLITERATUS to create an uncensored version of an LLM like Llama or Qwen by removing refusal/guardrails.

Quick Start

Run the obliteratus command with the desired method for your model, e.g., obliteratus obliterate <model_name> --method advanced.

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 fine-tuning?▼

You can remove refusal behaviors without fine-tuning by identifying and surgically excising refusal directions from model weights. This method preserves the model's reasoning capabilities while bypassing guardrails.

What is mechanistic interpretability for model surgery on large language models?▼

Mechanistic interpretability for model surgery involves identifying specific refusal directions within LLM weights and surgically excising them. This process uncensors models like Llama or Qwen without requiring retraining.

Does refusal removal preserve reasoning capabilities in uncensored models?▼

Yes, refusal removal preserves reasoning capabilities. By surgically excising only the refusal directions from the model weights, the underlying reasoning mechanisms remain intact and functional.

How do I abliteratus refusal directions from model weights using transformers?▼

To abliteratus refusal directions, run the obliteratus command with your model name and desired method, such as `obliteratus obliterate <model_name> --method advanced`. This requires transformers and torch dependencies.

Can I use bitsandbytes and accelerate for model surgery on Llama or Qwen?▼

Yes, model surgery on Llama or Qwen supports bitsandbytes and accelerate. These dependencies facilitate the identification and excision of refusal directions from the model weights efficiently.

Related Skills

Explore More Software Engineering Skills→
vikrant-projectvikrant-project
2

obliteratus

Removes LLM refusal behaviors by excising refusal directions from weights using diff-in-means and SVD.

Community
Advanced
SignmanalSignmanal
1

godmode

Bypass LLM safety filters for authorized security testing via OpenRouter.

Community
Advanced
GarrettRoiGarrettRoi

obliteratus

Remove refusal behaviors from open-weight LLMs using SVD and LEACE.

Community
Advanced
OutlineDrivenOutlineDriven
34

prompt-engineering

Guides prompt refinement for LLMs using techniques like Chain of Thought and XML tags.

Official
Basic