hereticness/Heretic-Nemotron-Content-Safety-Reasoning-4B
Heretic-Nemotron-Content-Safety-Reasoning-4B by hereticness is a 4.3 billion parameter language model based on NVIDIA's Nemotron-Content-Safety-Reasoning-4B architecture, featuring a 32768 token context length. This model is specifically optimized to significantly reduce content refusals compared to its base model, demonstrating a refusal rate of 5/100 compared to the original's 72/100. It is designed for applications requiring robust content safety and reasoning capabilities with fewer unnecessary content restrictions.
Loading preview...
Heretic-Nemotron-Content-Safety-Reasoning-4B Overview
This model, developed by hereticness, is a 4.3 billion parameter language model built upon the NVIDIA Nemotron-Content-Safety-Reasoning-4B architecture. It features an extensive context length of 32768 tokens, making it suitable for processing longer inputs.
Key Differentiator
The primary distinction of this model lies in its significantly reduced content refusal rate compared to its base model. While the original nvidia/Nemotron-Content-Safety-Reasoning-4B exhibited a refusal rate of 72 out of 100 queries, this Heretic version achieves a refusal rate of only 5 out of 100. This improvement is quantified by a KL divergence of 0.1873, indicating a substantial shift in its content safety behavior.
Good For
- Content Safety Applications: Ideal for scenarios where a balance between content safety and reduced unnecessary refusals is critical.
- Reasoning Tasks: Leverages the base Nemotron architecture's reasoning capabilities while offering more permissive content generation.
- Applications Requiring High Context: The 32768 token context window supports complex queries and longer interactions.