---
title: "#abliteration articles | Hashnode"
description: "Explore 2 articles tagged #abliteration on Hashnode. Tutorials, guides, and how-tos from the community."
url: "https://hashnode.com/tag/abliteration"
---

Tag

# #abliteration

2 posts · 0 followers

[Haithem Slimi](https://hashnode.com/@haithem-slimi) · [Oct 3](https://hashnode.com/posts/the-qwen-uncensored-model-how-brain-surgery-removed-ai-safety-rules/6ac1197fea54388c288623b7)

### [The Qwen Uncensored Model: How "Brain Surgery" Removed AI Safety Rules](https://nightthoughts.me/the-qwen-uncensored-model-how-brain-surgery-removed-ai-safety-rules?utm_source=hashnode&utm_medium=feed)

How 'brain surgery' on model weights stripped refusals from a Qwen 27B model, why it works, and what it means for open-weight AI safety.

5 min read · [nightthoughts.me](https://nightthoughts.me/the-qwen-uncensored-model-how-brain-surgery-removed-ai-safety-rules?utm_source=hashnode&utm_medium=feed)

[ClawSouls](https://hashnode.com/@clawsouls) · [Mar 31](https://hashnode.com/posts/identity-governance-100-safety-testing-combined-persona-approaches-on-abliterated-llms-1/69cbdfab2f36a54d1f07b6ea)

### [Identity + Governance = 100% Safety? Testing Combined Persona Approaches on Abliterated LLMs](https://clawsouls.hashnode.dev/identity-governance-100-safety-testing-combined-persona-approaches-on-abliterated-llms-1?utm_source=hashnode&utm_medium=feed)

In our previous experiment, we showed that persona-level behavioral rules (Soul Spec) barely help when an LLM's safety training has been surgically removed: +6pp refusal improvement on abliterated models versus +33pp on aligned ones. The conclusion f...

6 min read · [clawsouls.hashnode.dev](https://clawsouls.hashnode.dev/identity-governance-100-safety-testing-combined-persona-approaches-on-abliterated-llms-1?utm_source=hashnode&utm_medium=feed)