huihui-ai/Huihui-LFM2.5-1.2B-Instruct-abliterated
The huihui-ai/Huihui-LFM2.5-1.2B-Instruct-abliterated is a 1.2 billion parameter instruction-tuned causal language model, derived from LiquidAI/LFM2.5-1.2B-Instruct. This model has been specifically modified using an 'abliteration' technique to remove refusal behaviors, making it an uncensored version. It is designed for research and experimental use where reduced safety filtering is desired, offering a 32768 token context length.
Loading preview...
Overview
This model, huihui-ai/Huihui-LFM2.5-1.2B-Instruct-abliterated, is a 1.2 billion parameter instruction-tuned language model. It is an uncensored variant of the original LiquidAI/LFM2.5-1.2B-Instruct, created through an 'abliteration' process aimed at removing refusal behaviors. This proof-of-concept implementation focuses on demonstrating the removal of refusals without using TransformerLens.
Key Characteristics
- Uncensored Output: Safety filtering has been significantly reduced, allowing for potentially sensitive or controversial content generation.
- Experimental Focus: Primarily intended for research, testing, and controlled environments.
- Context Length: Supports a context window of 32768 tokens.
- Ollama Support: Available for direct use via Ollama, with a recommendation to use the bf16 version.
Usage Warnings & Considerations
Due to its modified nature, this model carries several important warnings:
- Risk of Sensitive Outputs: May generate inappropriate content; users must exercise caution and review outputs rigorously.
- Not for All Audiences: Unsuitable for public-facing applications or underage users.
- Legal & Ethical Responsibility: Users are solely responsible for ensuring compliance with laws and ethical standards.
- No Default Safety Guarantees: Unlike standard models, it lacks rigorous safety optimization, and the developers bear no responsibility for its use.