Open-Qwen2VL: Compute-Efficient Pre-Training of Fully-Open Multimodal LLMs on Academic Resources Paper • 2504.00595 • Published 6 days ago • 33 • 7