skills.rest
Brilly-Bohyun
Brilly-BohyunCOMMUNITY
Explore all skills by this author →
GitHub

obliteratus

Remove refusal behaviors from open-weight LLMs using diff-in-means, SVD, and SAE decomposition.

Updated Jun 17, 2026
Search Tags:#LLM#interpretability#refusal removal#model surgery#ablitation
One-click install
npx skills add https://github.com/Brilly-Bohyun/skill-repository --skill obliteratus-brilly-bohyun
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: obliteratus
Source: https://github.com/Brilly-Bohyun/skill-repository/tree/main/mlops/inference/obliteratus
Command: npx skills add https://github.com/Brilly-Bohyun/skill-repository --skill obliteratus-brilly-bohyun

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 resolves the issue of refusal behaviors in LLMs, allowing you to safely remove these guardrails without retraining or fine-tuning.

Core Features & Use Cases

  • Refusal Removal: Eliminate refusal behaviors (guardrails) from open-weight LLMs.
  • Mechanistic Interpretability: Utilizes advanced techniques like diff-in-means, SVD, and SAE decomposition for targeted removal.
  • Use Case: If you have a large LLM that exhibits refusal behaviors and you want to create a version that is less constrained, this skill can help you do so.

Quick Start

Run the obliteratus command to begin the ablitation process for your selected 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 without retraining by using mechanistic interpretability techniques like diff-in-means, SVD, and SAE decomposition to perform targeted ablitation on the model's specific refusal directions.

What is ablitation in the context of LLM interpretability?▼

Ablitation is a mechanistic interpretability technique that identifies and surgically removes specific refusal directions in LLMs using diff-in-means and SVD, eliminating guardrails without requiring full model retraining or fine-tuning.

Do I need to install PyTorch and transformers to remove LLM guardrails?▼

Yes, removing LLM guardrails requires installing torch, transformers, bitsandbytes, accelerate, and safetensors to execute the ablitation scripts that process and modify the model weights directly.

Can I reduce model guardrails on open-weight LLMs?▼

Yes, you can reduce guardrails on open-weight LLMs by applying targeted model surgery techniques. This process modifies the existing model weights directly to eliminate refusal behaviors without needing data-intensive fine-tuning.

What's the best way to modify LLM weights for refusal removal?▼

The best way to modify weights for refusal removal is using advanced mechanistic interpretability, specifically SAE decomposition and SVD, to isolate and neutralize the exact refusal behaviors while preserving the model's overall performance.

Related Skills

Explore More Data & Analytics Skills→
HuskyDannyHuskyDanny

obliteratus

Remove refusal behaviors from open-weight LLMs using mechanistic interpretability techniques.

Community
Advanced
vikrant-projectvikrant-project
2

obliteratus

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

Community
Advanced
HouseGarofaloHouseGarofalo
3

huggingface-transformers

Run local inference, embeddings, and fine-tuning with Hugging Face Transformers.

Official
Advanced
jonnabiojonnabio
6

fine-tuning-with-trl

Fine-tune language models with TRL methods for human preference alignment.

Community
Advanced