Delta-Vector/Control-Nanuq-8B

Delta-Vector/Control-Nanuq-8B is an 8 billion parameter language model, fine-tuned from LLaMA 3.1 8B Supernova. It is specifically designed to minimize narration and produce concise responses, making it suitable for applications requiring direct and brief outputs. The model incorporates DPO and KTO reinforcement learning to enhance coherence, prose, and creativity, while maintaining a focus on brevity. Its primary strength lies in generating 'short and sweet' interactions.