1 paper · 1 filter
Ming Li, Chenrui Fan, Yize Cheng +2
Large language models increasingly expose reasoning traces, yet their underlying cognitive structure and steps remain difficult to identify and analyze beyond surface-level statist…