Open-Qwen2VL: Compute-Efficient Pre-Training of Fully-Open Multimodal LLMs on Academic Resources Paper • 2504.00595 • Published 3 days ago • 26