4 ms·
LLMs are trained on texts reflecting real-life conversations. Saying “please” helps in those, hence it also tends to help in LLM inference, which follows patter
by layer8 2y ago
LLMs are trained on texts reflecting real-life conversations. Saying “please” helps in those, hence it also tends to help in LLM inference, which follows patterns that are prevalent in the training data.