harimuliawan999/Huihui-Qwen3.5-4B-abliterated

Model Overview

The harimuliawan999/Huihui-Qwen3.5-4B-abliterated is a 4.5 billion parameter language model derived from the Qwen3.5-4B base, fine-tuned by huihui-ai using TRL. A key characteristic of this model is its significantly reduced safety filtering, distinguishing it from standard models by allowing for a broader range of content generation without typical censorship constraints. The fine-tuning process focused exclusively on the text component, not the image part, and it maintains a 32768 token context length.

Key Characteristics

• Uncensored Output: Safety filtering has been substantially reduced, enabling the generation of sensitive or controversial content.
• Text-Only Processing: The fine-tuning was applied only to the text capabilities, not to any image processing components.
• Compatibility: Usable with ollama (v0.17.7 or later) and llama.cpp for conversion.

Usage Considerations

This model is primarily intended for research and experimental use in controlled environments. Users should be aware of the following:

• Risk of Inappropriate Content: Due to limited filtering, the model may produce sensitive, controversial, or inappropriate outputs.
• Not for Public/Production Use: It is not recommended for public-facing commercial applications or environments requiring strict content moderation.
• User Responsibility: Users are solely responsible for ensuring compliance with legal and ethical standards regarding generated content.
• No Safety Guarantees: Unlike standard models, this version has not undergone rigorous safety optimization, and the developers bear no responsibility for consequences arising from its use.