1 paper
Yanxing Guo, Zihao Zheng, Fangzhou Wu +4
Tackling complex reasoning tasks typically relies on massive monolithic LLMs, which suffer from severe computational redundancy. While task decomposition through structured pipelin…