When Parts Are Greater Than Sums: Individual LLM Components Can Outperform Full Models.

Ting-Yun Chang, Jesse Thomason, Robin Jia

Browse the full EMNLP paper archive.