# Academic Research Specialized

ProofGPT-v0.1 is a language model based on the GPT-NeoX architecture with 6.7 billion parameters, trained on the proof-pile dataset.

A 3B-parameter Q&A model fine-tuned from FineMath-Llama-3B, trained using the Numia-1.5 Q&A dataset

LlavaGuard is a safety assessment system based on a vision-language model, primarily used for safety classification and compliance checks of image content.

An instruction fine-tuned large language model based on DictaLM-2.0, specifically optimized for Hebrew and English chat scenarios

TheBeagle is a 7-billion-parameter model trained on The Bagel dataset, optimized with DPO (Direct Preference Optimization) and UNA (Unified Neural Architecture) techniques, demonstrating excellent performance in multi-task scenarios.

An Italian-adapted large language model based on LLaMA 2, focused on Italian text generation tasks