nguyenthilaitrieulong/DeepSeek-R1-Distill-Llama-70B-abliterated
The nguyenthilaitrieulong/DeepSeek-R1-Distill-Llama-70B-abliterated model is an uncensored 70 billion parameter Llama-based language model, derived from deepseek-ai/DeepSeek-R1-Distill-Llama-70B. It features a 32768-token context length and has been modified using 'abliteration' techniques to remove refusal behaviors. This model is specifically designed for use cases requiring an LLM that avoids content refusal, making it suitable for diverse conversational and generative tasks.
Loading preview...
Model Overview
This model, nguyenthilaitrieulong/DeepSeek-R1-Distill-Llama-70B-abliterated, is a 70 billion parameter language model based on the Llama architecture, with a substantial context length of 32768 tokens. It is a modified version of deepseek-ai/DeepSeek-R1-Distill-Llama-70B.
Key Differentiator
The primary distinction of this model is its "abliterated" nature, meaning it has undergone a process to remove refusal behaviors. This was achieved using a proof-of-concept implementation of techniques described in remove-refusals-with-transformers.
Usage Notes
Users might need to provide an example to guide the model if it initially struggles or refuses to respond, particularly if the "" token does not appear. An example provided in the original documentation is:
How many 'r' characters are there in the word "strawberry"?Deployment
This model is readily available for use with Ollama. You can run it directly using the command:
ollama run huihui_ai/deepseek-r1-abliterated:70bIntended Use Cases
This model is particularly suited for applications where an uncensored LLM is desired, and where the removal of refusal mechanisms is a critical requirement for generating responses across a broad range of prompts.