huihui-ai/DeepSeek-R1-Distill-Llama-8B-abliterated

Model Overview

The huihui-ai/DeepSeek-R1-Distill-Llama-8B-abliterated is an 8 billion parameter language model based on the deepseek-ai/DeepSeek-R1-Distill-Llama-8B architecture. Its primary distinguishing feature is the application of "abliteration" techniques, a proof-of-concept method to remove refusal behaviors and content filtering from the original model without relying on TransformerLens.

Key Characteristics

• Uncensored Output: Modified to bypass typical refusal mechanisms, providing direct responses.
• Abliteration Method: Utilizes a novel approach for refusal removal, detailed in the remove-refusals-with-transformers project.
• Llama-based Architecture: Inherits the foundational capabilities of the Llama model family.
• Context Length: Supports a substantial context window of 32768 tokens.

Usage Notes

Users may need to provide an example to guide the model if it initially refuses to respond or if specific formatting like "" is expected. For instance, providing a simple question-answer pair can help establish the desired interaction pattern.

Deployment

This model is available for use with Ollama under the name huihui_ai/deepseek-r1-abliterated:8b.