Open-Bee/

Bee-Training-Data-Stage2

[🏠 Homepage] [📖 Arxiv Paper] [🤗 Models & Datasets] [💻 Code]

We introduce Bee-8B, a new state-of-the-art, fully open 8B Multimodal Large Language Model (MLLM) designed to close the performance gap with proprietary models by focusing on data quality. Bee-8B is trained on our new Honey-Data-15M corpus, a high-quality supervised fine-tuning (SFT) dataset of approximately 15… See the full description on the dataset page: https://huggingface.co/datasets/Open-Bee/Bee-Training-Data-Stage2.