wexyyyyyy/gemma-3-1b-it-heretic
The wexyyyyyy/gemma-3-1b-it-heretic is a 1 billion parameter instruction-tuned causal language model, derived from Google's Gemma 3 family, specifically a decensored version of unsloth/gemma-3-1b-it. This model has been modified using the Heretic tool to significantly reduce refusals, demonstrating a refusal rate of 2/100 compared to the original's 99/100. It is designed for applications requiring less restrictive content generation, while maintaining the base Gemma 3 capabilities for text generation and understanding.
Loading preview...
Overview
wexyyyyy/gemma-3-1b-it-heretic is a 1 billion parameter instruction-tuned model based on Google's Gemma 3 architecture, specifically a decensored variant of unsloth/gemma-3-1b-it. This model was created using the Heretic v1.2.0 tool, which modifies the model's weights to reduce content refusal rates.
Key Capabilities
- Decensored Output: Significantly reduced refusal rate (2/100) compared to the original model (99/100), making it suitable for less restricted content generation.
- Gemma 3 Base: Inherits the core capabilities of the Gemma 3 family, including text generation, question answering, summarization, and reasoning.
- Multimodal Potential: While this specific variant is text-focused, the underlying Gemma 3 architecture supports text and image input, with a context window of 32K tokens for the 1B size.
- Lightweight: At 1 billion parameters, it is designed for deployment in resource-constrained environments.
Good For
- Use cases requiring a less restrictive or "decensored" language model.
- Applications where the original Gemma 3 instruction-tuned model's safety filters are too aggressive.
- Experimentation with model behavior modification and content generation without typical refusal mechanisms.
- Text generation tasks on devices with limited computational resources.