Andronovo/Huihui-CyberStrike-OffSec-35B-abliterated
Huihui-CyberStrike-OffSec-35B-abliterated is a 35.1 billion parameter uncensored causal language model developed by huihui-ai. It is an abliterated version of oyildirim/CyberStrike-OffSec-35B, specifically modified to remove refusal behaviors. This model is intended for research and experimental use in controlled environments where reduced safety filtering is desired, offering a proof-of-concept for refusal removal techniques.
Loading preview...
Overview
Andronovo/Huihui-CyberStrike-OffSec-35B-abliterated is a 35.1 billion parameter language model created by huihui-ai. It is an uncensored variant of the oyildirim/CyberStrike-OffSec-35B model, developed using an "abliteration" technique to remove refusal behaviors. This model serves as a proof-of-concept for methods to modify LLM responses without using TransformerLens, aiming to provide a model with significantly reduced safety filtering.
Key Characteristics
- Uncensored Nature: Modified to remove refusal mechanisms, allowing for a broader range of outputs.
- Abliteration Technique: Utilizes a specific method (referenced as
remove-refusals-with-transformers) for altering model behavior. - Base Model: Derived from
oyildirim/CyberStrike-OffSec-35B. - MTP File: Includes an MTP file from
Qwen/Qwen3.6-35B-A3B, primarily forllama.cppusage.
Usage Warnings
Due to its uncensored nature, this model comes with significant warnings:
- Risk of Sensitive Outputs: May generate controversial, inappropriate, or sensitive content.
- Not for All Audiences: Unsuitable for public-facing applications, underage users, or high-security environments.
- User Responsibility: Users are solely responsible for legal and ethical compliance of generated content.
- Research Use Only: Recommended for research, testing, or controlled environments, not for production.
- No Safety Guarantees:
huihui.aidisclaims responsibility for consequences arising from its use, as it lacks rigorous safety optimization.