ChiKoi7/Bellatrix-Tiny-3B-R1-Heretic
Bellatrix-Tiny-3B-R1-Heretic by ChiKoi7 is a 3.2 billion parameter, instruction-tuned causal language model based on the DeepSeek-R1 architecture, optimized for multilingual dialogue, agentic retrieval, and summarization tasks. This specific version has been decensored using the Heretic tool, significantly reducing refusals compared to the original Bellatrix-Tiny-3B-R1 model. It excels in following complex instructions and supporting conversations across multiple languages.
Loading preview...
Bellatrix-Tiny-3B-R1-Heretic Overview
Bellatrix-Tiny-3B-R1-Heretic is a 3.2 billion parameter, instruction-tuned language model developed by ChiKoi7. It is a decensored variant of the prithivMLmods/Bellatrix-Tiny-3B-R1 model, created using the Heretic v1.1.0 tool. This model is built upon a reasoning-based architecture derived from the DeepSeek-R1 synthetic dataset entries, utilizing an optimized transformer architecture.
Key Capabilities
- Decensored Output: Significantly reduced refusal rates (4/100) compared to the original Bellatrix-Tiny-3B-R1 (93/100), allowing for broader content generation.
- Multilingual Dialogue: Optimized for conversational use cases across various languages.
- Agentic Retrieval: Capable of intelligent information retrieval within dialogue or query-response systems.
- Summarization Tasks: Efficiently condenses large texts into concise summaries.
- Instruction Following: Designed to follow complex, context-aware instructions to produce precise outputs.
Good For
- Applications requiring a less restrictive language model for diverse content generation.
- Developing multilingual chatbots and conversational AI systems.
- Tasks involving intelligent information retrieval and summarization.
- Scenarios demanding precise output generation based on complex instructions.