1 paper · 1 filter
Ziyu Chen, Yilun Zhao, Chengye Wang +3
Constructing scientific multimodal document reasoning datasets for foundation model training involves an inherent trade-off among scale, faithfulness, and realism. To address this…