Advanced Natural-based interaction for the ITAlian language: LLaMAntino-3-ANITA
LLaMAntino-3-ANITA is a state-of-the-art Large Language Model fine-tuned for the Italian language. It combines Supervised Fine-tuning (SFT), QLoRA for efficiency, and Dynamic Preference Optimization (DPO) to improve performance and alignment. The model shows outstanding results on Italian benchmarks, outperforming similar models in both accuracy and computational efficiency.