1 paper
Archiki Prasad, Mandar Joshi, Kenton Lee +2
Chain-of-thought (CoT) reasoning and its variants have substantially improved the performance of language models on complex reasoning tasks, yet the precise mechanisms by which dif…