Hugging Face daily papers·19d agoJust MLPs: Efficient Visual State Reconstruction for Multimodal Language Models#mllm#visual-tokens#efficiency