Harness Engineering：有時候語言模型不是不夠聰明，只是沒有人類好好引導

Lecture on harness engineering, showing that language models often underperform not from lack of capability but from insufficient human guidance