richardyoung/Llama-3.1-8B-Instruct-abliterated-obliteratus
The richardyoung/Llama-3.1-8B-Instruct-abliterated-obliteratus is an 8 billion parameter instruction-tuned causal language model, derived from Meta's Llama-3.1-8B-Instruct. Developed by Richard Young, this model has undergone an "abliteration" process using the OBLITERATUS method to significantly reduce refusal behavior. It is specifically designed for research into uncensored model responses, offering a 32768 token context length.
Loading preview...
Model Overview
This model, richardyoung/Llama-3.1-8B-Instruct-abliterated-obliteratus, is an 8 billion parameter instruction-tuned language model based on Meta's Llama-3.1-8B-Instruct. Its primary distinction is the application of an advanced "abliteration" technique, specifically the OBLITERATUS method, to remove refusal behaviors.
Key Characteristics
- Abliteration: Achieves a high refusal rate reduction (95/100) with a low Attack Success Rate (ASR) of 5.0%, indicating effective removal of safety guardrails.
- Research Focus: Developed as part of research into LLM abliteration methods, detailed in the paper "Comparative Analysis of LLM Abliteration Methods: Scaling to MoE Architectures and Modern Tools" by Richard Young.
- Context Length: Supports a substantial context window of 32768 tokens.
- Disclaimer: Released for research purposes only, with explicit warning that safety guardrails have been removed.
Use Cases
- Research into Model Alignment: Ideal for studying the effects of removing refusal mechanisms and exploring the underlying principles of model safety.
- Uncensored Content Generation: Suitable for specific research scenarios requiring models without built-in refusal behaviors, provided users adhere to ethical guidelines.
- Comparative Analysis: Can be used to compare abliteration techniques against other methods for modifying model behavior.