1 paper
Xiaoyang Chen, Jie Liu, Haijin Liang +5
In pointwise document reranking, Chain-of-Thought models typically underperform direct scoring models. While existing diagnostics attribute this to inferior classification, score p…