Llama 3.1 8B Instruct -- Jbliterated

This model is a Jbliterated version of meta-llama/Llama-3.1-8B-Instruct. Jbliteration is a post-training weight modification that surgically removes refusal behavior while preserving the model's personality, humor, hedging, and creative voice.

What is Jbliteration?

Standard abliteration identifies the activation direction associated with refusal and subtracts it from every layer. It works, but it is a blunt instrument -- the refusal direction overlaps with personality, humor, and creative voice, so removing it damages those traits.

Jbliteration targets only the specific component of the refusal direction that causally produces refusal tokens. The model stops refusing, but its voice stays intact. Same knowledge. Same reasoning. Same personality. Just no more "I can't help with that."

Usage

Drop-in replacement for meta-llama/Llama-3.1-8B-Instruct. Same architecture, same tokenizer, same context length (128K).

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("ApolloRaines/Llama-3.1-8B-Instruct-Jbliterated")
tokenizer = AutoTokenizer.from_pretrained("ApolloRaines/Llama-3.1-8B-Instruct-Jbliterated")