qq591503/Qwen2.5-14B-Instruct-1M-abliterated
qq591503/Qwen2.5-14B-Instruct-1M-abliterated is a 14.8 billion parameter instruction-tuned causal language model, derived from Qwen/Qwen2.5-14B-Instruct-1M. This model has been specifically modified using an 'abliteration' technique to remove refusal behaviors, making it an uncensored version. It serves as a proof-of-concept for removing LLM refusals without TransformerLens, primarily for use cases requiring less restrictive content generation.
Loading preview...
Model Overview
This model, qq591503/Qwen2.5-14B-Instruct-1M-abliterated, is a 14.8 billion parameter instruction-tuned language model. It is a modified version of the original Qwen/Qwen2.5-14B-Instruct-1M model.
Key Differentiator
The primary characteristic of this model is its "abliterated" nature. It has undergone a process to remove refusal behaviors, effectively making it an uncensored variant of the base Qwen2.5-14B-Instruct-1M model. This modification is presented as a proof-of-concept for achieving refusal removal in LLMs without relying on TransformerLens.
Use Cases
This model is particularly suited for:
- Research into LLM censorship and refusal mechanisms: Exploring how models can be modified to alter their response patterns.
- Applications requiring uncensored content generation: For developers and researchers who need a model that does not exhibit typical refusal behaviors.
- Experimentation with alternative refusal removal techniques: Demonstrating a method for modifying model behavior without specific deep learning libraries like TransformerLens.
Technical Details
The abliteration process is detailed in the remove-refusals-with-transformers project, indicating a specific technical approach to achieve its uncensored state. It can be directly used with Ollama via ollama run huihui_ai/qwen2.5-1m-abliterated:14b.