rahul7star/gemma-3-12b-it-heretic
The rahul7star/gemma-3-12b-it-heretic model is a 12 billion parameter instruction-tuned multimodal language model, based on Google's Gemma 3 architecture. This version has been 'decensored' using the Heretic v1.0.0 tool, significantly reducing refusals compared to the original google/gemma-3-12b-it model. It supports both text and image inputs with a 128K context window and excels in various text generation, image understanding, and reasoning tasks.
Loading preview...
Model Overview
This model, rahul7star/gemma-3-12b-it-heretic, is a modified version of Google's Gemma 3 12B instruction-tuned multimodal language model. It has been 'decensored' using the Heretic v1.0.0 tool, resulting in a substantial reduction in refusal rates (3/100 compared to 97/100 for the original model), as indicated by KL divergence metrics.
Key Capabilities
- Multimodal Input: Handles both text and image inputs, with images normalized to 896x896 resolution and encoded to 256 tokens each.
- Large Context Window: Features a 128K token total input context, enabling processing of extensive information.
- Multilingual Support: Supports over 140 languages.
- Reduced Refusals: Demonstrates significantly fewer refusals in responses compared to the base Gemma 3 12B IT model.
- Diverse Task Performance: Well-suited for a variety of tasks including question answering, summarization, reasoning, and image analysis.
Good For
- Applications requiring less restrictive content generation: Ideal for use cases where the original model's safety filters might be overly cautious.
- Text Generation: Creative writing, code generation, marketing copy, and email drafts.
- Conversational AI: Powering chatbots and virtual assistants.
- Image Understanding: Extracting, interpreting, and summarizing visual data.
- Research and Development: Serving as a foundation for VLM and NLP research, especially for exploring model behavior with reduced inherent censorship.